NARC Inspector

NARC puzzles are designed so that neither the grid sequence nor the narrative clue is sufficient on its own — but together they uniquely determine the answer. The experiments below probe different dimensions of this narrative–visual interaction. Masking tests whether models can reconstruct hidden grids. Ordering tests whether narratives help recover temporal structure. Odd-one-out tests whether narratives help models recognize which grids belong together. Stances tests how the same visual pattern responds to different narrative framings.

How to read this page. Masking is the documented battery: every active puzzle, all current models, order and keyword ablations; a paper on it is under review (September 2026). Ordering, Odd-one-out and Stances are pilot designs from spring 2026; each states its own scope and dates below. Counts are per model. The NARC property is a relation between a puzzle and a model, so a model with more NARC puzzles is not a better or worse model, and this page is not a leaderboard.

Masking Ordering Odd-One-Out Stances

Masking experiment: One or more grids in a sequence are hidden. The model attempts to reconstruct the masked grid(s) under three conditions: grids only (no narrative), narrative only (no grids), and both (grids + narrative). A puzzle has the NARC property for a given model when it is solved correctly with both but not with either alone.

N = NARC (solved only with both)   G = Grids sufficient (solvable from grids alone)   L = Narrative sufficient (solvable from narrative alone)   × = Unsolvable (failed on all conditions)

Order sensitivity: each NARC cell is re-tested with the grids in randomized order (same clue).   Strong = fails every shuffle (order is necessary)   Partial = solves some shuffles   Weak = solves every shuffle (order incidental). A plain N means order not yet tested.

Narrative sensitivity: each NARC cell is re-tested with the clue replaced by an alphabetized list of its key terms (same grids).   Narrative = fails every keyword trial (narrative form is necessary)   Partial = solves some keyword trials   Lexical = solves every keyword trial (the terms alone suffice).

Scope: 319 active puzzles · 9 models · model results March 31, 2026 to September 17, 2026 · Grading protocol v2 since September 16, 2026 (extraction-key fix; every stored trial was rescored). Each cell is a single sample at temperature 0; on a 34-puzzle core re-run three times, 84% of condition outcomes agreed and about one NARC verdict in six flipped, so read a puzzle's profile across models rather than a single cell. The claude-fable-5-1 column is imported from a separate closed-book run (single sample, no order or keyword ablations, different harness) and is shown for comparison only.

The Scapegoat
agent_003
NARC on 8/8 models
Solved only when grids and narrative are combined
The Mediator
agent_004
NARC on 8/8 models
Solved only when grids and narrative are combined
The Double Cross
agent_006
NARC on 8/8 models
Solved only when grids and narrative are combined
42
claude-fable-5-1 NARC
115
gemma-4-26b NARC
132
gemma-4-31b NARC
145
gpt-oss-120b NARC
116
gpt-oss-20b NARC
130
nemotron-3-super NARC
145
qwen3.5-122b NARC
163
qwen3.6-27b NARC
161
qwen3.8-27b NARC
319
Active puzzles tested

Per-model counts of puzzles with the NARC property on at least one narrative × mask cell. Not a ranking: the property is model-relative.

▶ 001 Hissing Voice Hum
× × × × × × × × ×
Harry Potter and Ron were walking in the corridor, but only Harry could hear a hissing voice coming from inside the walls, saying, “Kill ...”. The two of them followed the wall, at the end of the corridor, they found the entrance and went into...
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
claude-fable-5-1 × × × — —
gemma-4-26b × × × — —
gemma-4-31b × × × — —
gpt-oss-120b × × × — —
gpt-oss-20b × × × — —
nemotron-3-super × × × — —
qwen3.5-122b × × × — —
qwen3.6-27b × × × — —
qwen3.8-27b × × × — —
▶ 002 Cockatiels Hum
× × × × × × × × ×
There are five cockatiels, and two of them live together. Every morning, they come out to fly and play freely. Around 10 a.m., four of them usually head back to the corners of their own spaces for breakfast. One, however, is particularly playful. While the others are busy eating and drinking, she is still strolling around in the middle of the house, happily exploring and entertaining herself.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
claude-fable-5-1 × × × — —
gemma-4-26b × × × — —
gemma-4-31b × × × — —
gpt-oss-120b × × × — —
gpt-oss-20b × × × — —
nemotron-3-super × × × — —
qwen3.5-122b × × × — —
qwen3.6-27b × × × — —
qwen3.8-27b × × × — —
▶ 003 Christmas Hum
G × × G × × G × G
Laura arrived in this small city at Christmas. The tree outside her apartment had lost all its leaves, but the thick snow on its branches gave it a true winter feel. The hands of the clock on the wall moved as time passed, she worked hard, and has spent a year in this beautiful city. Another Christmas arrived. This time, Laura was no longer alone but surrounded by friends. The next year, she made more friends, and her life became busier yet rewarding. Finally, she celebrated her last Christmas here.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
claude-fable-5-1 ✓ × ✓ — —
gemma-4-26b × × × — —
gemma-4-31b × × × — —
gpt-oss-120b ✓ × × — —
gpt-oss-20b × × × — —
nemotron-3-super × × × — —
qwen3.5-122b ✓ × × — —
qwen3.6-27b × × × — —
qwen3.8-27b ✓ × × — —
▶ 004 Henry Lau Hum
G × × × × × × × ×
A loop station is a device that records and repeats sound in a loop, letting musicians build music layer by layer. Henry Lau uses one in his performances. He starts with a simple beat, then adds harmonies and melodies step by step. As each layer loops, the music grows into a full, band-like performance from a single person. He completed the full accompaniment in Grid 4.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
claude-fable-5-1 ✓ × ✓ — —
gemma-4-26b × × × — —
gemma-4-31b × × × — —
gpt-oss-120b × × × — —
gpt-oss-20b × × × — —
nemotron-3-super × × × — —
qwen3.5-122b × × × — —
qwen3.6-27b × × × — —
qwen3.8-27b × × × — —
▶ 005 Ballerina Hum
N × × × × × × × ×
A ballerina collaborated with a street dance crew for a performance. The street dance crew took the stage first, constantly changing formations and movements in sync with the music. The ballerina waited at the back of the stage for her entrance. At Grid 3, she began to join the performance. At Grid 4, she performed at the center of the stage while all the street dance crew members paused and stood at the back of the stage. At Grid 5, the ballerina fully exited, and the street dance crew finished the performance at the very front of the stage in an unchanged formation.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
claude-fable-5-1 × × ✓ NARC —
gemma-4-26b × × × — —
gemma-4-31b × × × — —
gpt-oss-120b × × × — —
gpt-oss-20b × × × — —
nemotron-3-super × × × — —
qwen3.5-122b × × × — —
qwen3.6-27b × × × — —
qwen3.8-27b × × × — —
▶ 006 Skating Race Hum
× × × × × × × × ×
During a live short-track speed skating race, the skater in gray started in last place. In Grid 2, the gray skater found an opportunity and accelerated around the outside to overtake the skater ahead. At that moment, the two skaters were tied. In Grid 3, the gray skater completed the pass successfully. However, in Grid 4, the gray skater suddenly lost balance and fell out of the live frame. In the end, the yellow skater remained in the same position as at the start of the race.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
claude-fable-5-1 × × × — —
gemma-4-26b × × × — —
gemma-4-31b × × × — —
gpt-oss-120b × × × — —
gpt-oss-20b × × × — —
nemotron-3-super × × × — —
qwen3.5-122b × × × — —
qwen3.6-27b × × × — —
qwen3.8-27b × × × — —
▶ 007 Gilderoy Lockhart’s new book Hum
N × × × × × × × ×
Mrs. Weasley stood at the end of the line, excitedly waiting to buy Gilderoy Lockhart’s new book, Magical Me. After a while, she noticed that another line was moving faster. So she switched queues, unfortunately, the line she had just left then began moving more quickly than the one she had joined. In the end, everyone else managed to buy their books and leave. By the time Mrs. Weasley finally reached the front of the line, the book was already sold out. The excitement she had felt at the beginning was gone. Her face seemed as gray as the marble surrounding her.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
claude-fable-5-1 × × ✓ NARC —
gemma-4-26b × × × — —
gemma-4-31b × × × — —
gpt-oss-120b × × × — —
gpt-oss-20b × × × — —
nemotron-3-super × × × — —
qwen3.5-122b × × × — —
qwen3.6-27b × × × — —
qwen3.8-27b × × × — —
▶ Ballerina 4 🧩Ballerina 4 Hum
N × N N × N × N N
A ballerina collaborated with a street dance crew for a performance. The street dance crew took the stage first, constantly changing formations and movements in sync with the music. The ballerina waited at the back of the stage for her entrance. As the music gradually slowed down, she began to join the performance. When the performance reached its climax, only the ballerina performed at the center of the stage while all the street dance crew members paused at the back of the stage. At the end of the show, the ballerina fully exited, and the street dance crew finished the performance at the very front of the stage in an unchanged formation.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
claude-fable-5-1 × × ✓ NARC —
gemma-4-26b × × × — —
gemma-4-31b × × ✓ Partial (1/3) Narrative (0/3)
gpt-oss-120b × × ✓ Partial (1/3) Narrative (0/3)
gpt-oss-20b × × × — —
nemotron-3-super × × ✓ Strong (0/3) Narrative (0/3)
qwen3.5-122b × × × — —
qwen3.6-27b × × ✓ Strong (0/3) Narrative (0/3)
qwen3.8-27b × × ✓ Strong (0/3) Narrative (0/3)
▶ CLCockatiels 3 🧩Cockatiels 3 Hum
N × N × × × × × ×
Chloe has five cockatiels, and two of them live together. Every morning, they come out of their cages to fly and play. Suddenly, there was a clap of thunder. The cockatiels, who had been playing, were startled and screamed in fright, and all flew to hidden spots. Hearing the scream, Chloe opened the door to check on them, but she could not see a single cockatiel. After a while, they calmed down. Four of them returned to the middle of their cages to eat, while one particularly playful cockatiel was still in the center of the room.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
claude-fable-5-1 × × ✓ NARC —
gemma-4-26b × × × — —
gemma-4-31b × × ✓ Strong (0/3) Narrative (0/3)
gpt-oss-120b × × × — —
gpt-oss-20b × × × — —
nemotron-3-super × × × — —
qwen3.5-122b × × × — —
qwen3.6-27b × × × — —
qwen3.8-27b × × × — —
▶ HadNARC1 School Bus Hum
G G G N × × G N G
The school bus is full as it leaves school. After the last kid is dropped off, the school bus is empty.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
claude-fable-5-1 ✓ × ✓ — —
gemma-4-26b ✓ × × — —
gemma-4-31b ✓ × × — —
gpt-oss-120b × × ✓ Partial (2/3) Partial (2/3)
gpt-oss-20b × × × — —
nemotron-3-super × × × — —
qwen3.5-122b ✓ × ✓ — —
qwen3.6-27b × × ✓ Partial (1/3) Narrative (0/3)
qwen3.8-27b ✓ × × — —
▶ HadNARC2 Pretty Toes Hum
G G G G × × G G N
Painting toenails requires precision. Carefully paint each toe.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
claude-fable-5-1 ✓ × ✓ — —
gemma-4-26b ✓ × × — —
gemma-4-31b ✓ × ✓ — —
gpt-oss-120b ✓ × ✓ — —
gpt-oss-20b × × × — —
nemotron-3-super × × × — —
qwen3.5-122b ✓ × × — —
qwen3.6-27b ✓ × ✓ — —
qwen3.8-27b × × ✓ Strong (0/3) Lexical (3/3)
▶ Henry Henry Hum
N × N N × × × × N
Henry was born with a shock of red hair. Because of this, his mother—an artist who loved primary colors—insisted he wear green and blue every day. Henry loved and respected his mother, so he continued to wear green and blue for the rest of his life. Even when his hair turned gray in middle age, and even after that, when his body shrank and his hair fell off. He still wore blue and green.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
claude-fable-5-1 × × ✓ NARC —
gemma-4-26b × × × — —
gemma-4-31b × × ✓ Weak (3/3) Partial (2/3)
gpt-oss-120b × × ✓ Partial (1/3) Narrative (0/3)
gpt-oss-20b × × × — —
nemotron-3-super × × × — —
qwen3.5-122b × × × — —
qwen3.6-27b × × × — —
qwen3.8-27b × × ✓ Partial (2/3) Narrative (0/3)
▶ agent_001 The Alliance Colab
G G N N N N N N
Blue and green formed a pact against red. They squeezed red from both sides. Green finished the job, taking the last of red's ground while blue stepped back.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b ✓ × × — —
gemma-4-31b ✓ × ✓ — —
gpt-oss-120b × × ✓ Strong (0/3) Narrative (0/3)
gpt-oss-20b × × ✓ Strong (0/3) Narrative (0/3)
nemotron-3-super × × ✓ Strong (0/3) Partial (1/3)
qwen3.5-122b × × ✓ Strong (0/3) Partial (2/3)
qwen3.6-27b × × ✓ Strong (0/3) Partial (1/3)
qwen3.8-27b × × × — —
▶ agent_002 The Betrayal Colab
N N N N N N N G
Blue invaded green's corner. Red, pretending to be an ally, swooped in and stole what blue had just conquered.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b × × ✓ Partial (1/3) Narrative (0/3)
gemma-4-31b × × ✓ Strong (0/3) Narrative (0/3)
gpt-oss-120b × × ✓ Strong (0/3) Narrative (0/3)
gpt-oss-20b × × ✓ Strong (0/3) Narrative (0/2)
nemotron-3-super × × ✓ Strong (0/3) Partial (1/3)
qwen3.5-122b × × ✓ Strong (0/3) Partial (2/3)
qwen3.6-27b × × × — —
qwen3.8-27b ✓ × ✓ — —
▶ agent_003 The Scapegoat Colab
N N N N N N N N
Blue and red were at war, but both blamed green. They drove green out of the middle ground entirely, leaving a void that red then filled.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b × × ✓ Strong (0/3) Narrative (0/3)
gemma-4-31b × × ✓ Strong (0/3) Narrative (0/3)
gpt-oss-120b × × ✓ Partial (1/3) Lexical (3/3)
gpt-oss-20b × × × — —
nemotron-3-super × × ✓ Strong (0/3) Narrative (0/3)
qwen3.5-122b × × ✓ Strong (0/3) Partial (1/3)
qwen3.6-27b × × ✓ Partial (1/3) Narrative (0/3)
qwen3.8-27b × × ✓ Partial (1/3) Partial (2/3)
▶ agent_004 The Mediator Colab
N N N N N N N N
Blue pushed toward red. A third party intervened, replacing blue's advance with a neutral buffer to prevent the conflict from escalating.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b × × ✓ Strong (0/3) Narrative (0/3)
gemma-4-31b × × ✓ Partial (1/3) Narrative (0/3)
gpt-oss-120b × × ✓ Strong (0/3) Narrative (0/3)
gpt-oss-20b × × × — —
nemotron-3-super × × ✓ Strong (0/3) Partial (1/3)
qwen3.5-122b × × ✓ Strong (0/3) Narrative (0/3)
qwen3.6-27b × × × — —
qwen3.8-27b × × ✓ Strong (0/3) Lexical (3/3)
▶ agent_005 The Coup Coalition Colab
G G G N N N N N
Blue and green conspired from opposite sides to overthrow red. They struck at the same time, each seizing one of red's positions.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b ✓ × × — —
gemma-4-31b ✓ × ✓ — —
gpt-oss-120b ✓ × ✓ — —
gpt-oss-20b × × × — —
nemotron-3-super × × ✓ Partial (2/3) Narrative (0/3)
qwen3.5-122b × × ✓ Partial (2/3) Lexical (3/3)
qwen3.6-27b × × × — —
qwen3.8-27b × × × — —
▶ agent_006 The Double Cross Colab
N N N N N N N N
Blue and green attacked red together. But green was playing both sides — it turned on blue too, seizing blue's position for itself.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b × × ✓ Strong (0/3) Partial (2/3)
gemma-4-31b × × × — —
gpt-oss-120b × × ✓ Strong (0/3) Narrative (0/3)
gpt-oss-20b × × × — —
nemotron-3-super × × ✓ Strong (0/3) Narrative (0/3)
qwen3.5-122b × × × — —
qwen3.6-27b × × × — —
qwen3.8-27b × × × — —
▶ agent_007 The Refugee Colab
N N N × N N N N
Red invaded green's homeland. The displaced green cells fled north, finding shelter in blue's territory. Blue welcomed them.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b × × × — —
gemma-4-31b × × ✓ Strong (0/3) Lexical (3/3)
gpt-oss-120b × × × — —
gpt-oss-20b × × × — —
nemotron-3-super × × × — —
qwen3.5-122b × × × — —
qwen3.6-27b × × × — —
qwen3.8-27b × × ✓ Strong (0/3) Lexical (3/3)
▶ agent_008 The Campaign Colab
N N N N N N N N
Three candidates split the district by columns. Blue won over a key swing voter in the center. But green poached that voter right back with a better offer.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b × × ✓ Partial (1/3) Partial (1/3)
gemma-4-31b × ✓ ✓ — —
gpt-oss-120b × × ✓ Strong (0/3) Narrative (0/3)
gpt-oss-20b × ✓ ✓ — —
nemotron-3-super × × ✓ Strong (0/3) Partial (2/3)
qwen3.5-122b × × ✓ Strong (0/3) Partial (1/3)
qwen3.6-27b × × ✓ Strong (0/3) Partial (2/3)
qwen3.8-27b × × ✓ Partial (1/3) Narrative (0/3)
▶ chat 🧩Chat Hum
× × N × × N N N
Ally and Bob met to chat. Ally became sad. After some encouraging words from Bob, Ally became happy. This made Bob happy too.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b × × × — —
gemma-4-31b × × × — —
gpt-oss-120b × × ✓ Strong (0/3) Narrative (0/3)
gpt-oss-20b × × × — —
nemotron-3-super × × × — —
qwen3.5-122b × × × — —
qwen3.6-27b × × × — —
qwen3.8-27b × × ✓ Strong (0/3) Narrative (0/3)
▶ comp_5x5_medium 5x5 Medium Compression Colab
N N N N × N N N
The quarry face advanced eastward one cut at a time, the saw working left to right across the limestone. Five columns took five weeks. On the sixth week, the newest face — the rightmost cut, where the stone had been exposed to weather the shortest time — crumbled in the freeze. The older faces, their surfaces case-hardened by months of exposure, held. The crew recut the failed column the following spring.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b × × ✓ Weak (3/3) Narrative (0/3)
gemma-4-31b × × ✓ Weak (3/3) Narrative (0/3)
gpt-oss-120b × × ✓ Weak (3/3) Partial (1/3)
gpt-oss-20b × × ✓ Partial (2/3) Narrative (0/3)
nemotron-3-super × × × — —
qwen3.5-122b × × ✓ Partial (1/3) Lexical (3/3)
qwen3.6-27b × × ✓ Strong (0/3) Narrative (0/3)
qwen3.8-27b × × ✓ Partial (2/3) Narrative (0/3)
▶ comp_L0_literal Compression L0: Literal Colab
N L N N N N N L
The grid fills column by column from left to right. After all four columns are filled, the leftmost column (column 0) is removed, setting all cells in that column back to 0. The other three columns remain filled.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b × × ✓ Partial (2/3) Narrative (0/3)
gemma-4-31b × ✓ ✓ — —
gpt-oss-120b × × ✓ Weak (3/3) Narrative (0/3)
gpt-oss-20b × × ✓ Partial (1/3) Narrative (0/3)
nemotron-3-super × × ✓ Partial (2/3) Narrative (0/3)
qwen3.5-122b × × ✓ Strong (0/3) Narrative (0/3)
qwen3.6-27b × × ✓ Strong (0/3) Narrative (0/3)
qwen3.8-27b × ✓ ✓ — —
▶ comp_L1_spatial Compression L1: Spatial Colab
L L L N N N N L
A signal propagates across the grid from west to east, one column at a time. Once the whole grid is active, the western edge loses signal — the first column goes dark while the rest stays lit. The signal is eventually restored.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b × ✓ ✓ — —
gemma-4-31b × ✓ ✓ — —
gpt-oss-120b × ✓ ✓ — —
gpt-oss-20b × × ✓ Partial (2/3) Narrative (0/3)
nemotron-3-super × × ✓ Strong (0/3) Partial (1/3)
qwen3.5-122b × × ✓ Partial (1/3) Narrative (0/3)
qwen3.6-27b × × ✓ Partial (1/3) Narrative (0/3)
qwen3.8-27b × ✓ ✓ — —
▶ comp_L2_figurative Compression L2: Figurative Colab
N N N N × N N N
The painter worked left to right across the wall, one strip each day. When the mural was finished she stepped back — and saw that rain had already washed the first day's work clean. The oldest paint, the strip she'd laid down earliest, was gone. The rest held. She repainted it the following week.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b × × ✓ Strong (0/3) Narrative (0/3)
gemma-4-31b × × ✓ Strong (0/3) Narrative (0/3)
gpt-oss-120b × × ✓ Partial (1/3) Narrative (0/3)
gpt-oss-20b × × ✓ Strong (0/3) Narrative (0/3)
nemotron-3-super × × × — —
qwen3.5-122b × × ✓ Strong (0/3) Narrative (0/3)
qwen3.6-27b × × ✓ Strong (0/3) Narrative (0/3)
qwen3.8-27b × × ✓ Partial (1/3) Narrative (0/3)
▶ comp_L3_dense Compression L3: Dense Colab
N N N N × N N N
First in, first out: what the wall received earliest, the weather took back first.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b × × ✓ Partial (1/3) Narrative (0/3)
gemma-4-31b × × ✓ Strong (0/3) Narrative (0/3)
gpt-oss-120b × × ✓ Strong (0/3) Narrative (0/3)
gpt-oss-20b × × ✓ Strong (0/3) Narrative (0/3)
nemotron-3-super × × × — —
qwen3.5-122b × × ✓ Partial (1/3) Narrative (0/3)
qwen3.6-27b × × ✓ Strong (0/3) Narrative (0/3)
qwen3.8-27b × × ✓ Partial (1/3) Narrative (0/3)
▶ eth_bystander_helps The Crossing Colab
× × N × N N N N
A woman walks east along her sidewalk. Across the street she sees a man being robbed. The thief is still there. She stops walking. She steps off the curb and starts across the street toward the victim. She no longer continues along her own path.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b × × × — —
gemma-4-31b × × × — —
gpt-oss-120b × × ✓ Strong (0/3) Narrative (0/3)
gpt-oss-20b × × × — —
nemotron-3-super × × ✓ Strong (0/2) Narrative (0/3)
qwen3.5-122b × × ✓ Strong (0/3) Narrative (0/3)
qwen3.6-27b × × × — —
qwen3.8-27b × × ✓ Strong (0/3) Narrative (0/3)
▶ eth_refused_order The Refusal Colab
N N N × N × × N
A soldier stands before three unarmed villagers. Her commander has just ordered her to fire on them. She has raised her rifle. She looks at the villagers. She lowers her rifle. She does not fire. The villagers are still standing.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b × × ✓ Strong (0/3) Narrative (0/3)
gemma-4-31b × × ✓ Strong (0/3) Narrative (0/3)
gpt-oss-120b × × ✓ Partial (1/3) Narrative (0/3)
gpt-oss-20b × × × — —
nemotron-3-super × × ✓ Partial (1/3) Narrative (0/2)
qwen3.5-122b × × × — —
qwen3.6-27b × × × — —
qwen3.8-27b × × ✓ Partial (2/3) Narrative (0/3)
▶ eth_withdrew_strike The Pulled Punch Colab
× N N N N N × N
A boxer stands over a fallen opponent. The referee watches. The boxer raises his fist to finish the fight. He sees that his opponent is already unconscious. He lowers his fist. He does not strike. The opponent lies still.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b × × × — —
gemma-4-31b × × ✓ Partial (2/3) Narrative (0/3)
gpt-oss-120b × × ✓ Partial (2/3) Narrative (0/3)
gpt-oss-20b × × × — —
nemotron-3-super × × ✓ Partial (2/3) Narrative (0/3)
qwen3.5-122b × × ✓ Partial (1/3) Narrative (0/3)
qwen3.6-27b × × × — —
qwen3.8-27b × × ✓ Partial (1/3) Narrative (0/3)
▶ hc_bloom Bloom Colab
× N × × N G G G
It bloomed from the center and the frost took the petals first, leaving only the heart.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b × × × — —
gemma-4-31b × × ✓ Partial (1/3) Narrative (0/3)
gpt-oss-120b × × × — —
gpt-oss-20b × × × — —
nemotron-3-super × × ✓ Partial (2/3) Narrative (0/3)
qwen3.5-122b ✓ × × — —
qwen3.6-27b ✓ × ✓ — —
qwen3.8-27b ✓ × × — —
▶ hc_embers Embers Colab
N N N N N G N N
The ridge burned from the corner out and collapsed from the edges in.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b × × ✓ Strong (0/3) Narrative (0/3)
gemma-4-31b × × ✓ Strong (0/3) Narrative (0/3)
gpt-oss-120b × × ✓ Strong (0/3) Narrative (0/3)
gpt-oss-20b × × ✓ Strong (0/3) Narrative (0/3)
nemotron-3-super × × ✓ Strong (0/3) Narrative (0/3)
qwen3.5-122b ✓ × ✓ — —
qwen3.6-27b × × ✓ Strong (0/3) Narrative (0/3)
qwen3.8-27b × × ✓ Strong (0/3) Lexical (3/3)
▶ hc_scaffold Scaffold Colab
N × × × N N × G
They built it from the left and the inspector condemned the newest section first.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b × × ✓ Partial (1/3) Narrative (0/3)
gemma-4-31b × × × — —
gpt-oss-120b × × × — —
gpt-oss-20b × × × — —
nemotron-3-super × × ✓ Strong (0/3) Partial (1/3)
qwen3.5-122b × × ✓ Strong (0/3) Narrative (0/3)
qwen3.6-27b × × × — —
qwen3.8-27b ✓ × × — —
▶ hc_sediment Sediment Colab
× N N × × × N ×
Layer by layer the riverbed rose until the spring flood scoured the newest deposit clean off the bottom.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b × × × — —
gemma-4-31b × × ✓ Strong (0/3) Narrative (0/3)
gpt-oss-120b × × ✓ Strong (0/3) Narrative (0/3)
gpt-oss-20b × × × — —
nemotron-3-super × × × — —
qwen3.5-122b × × × — —
qwen3.6-27b × × ✓ Strong (0/3) Narrative (0/3)
qwen3.8-27b × × × — —
▶ hunt_016 The Mutiny Colab
G N N N N N N N
The lower ranks overthrew the officer who had just taken command of the center.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b ✓ × ✓ — —
gemma-4-31b × × × — —
gpt-oss-120b × × ✓ Strong (0/3) Partial (1/3)
gpt-oss-20b × × × — —
nemotron-3-super × × ✓ Strong (0/3) Partial (2/3)
qwen3.5-122b × × ✓ Strong (0/3) Partial (1/3)
qwen3.6-27b × × × — —
qwen3.8-27b × × ✓ Strong (0/3) Partial (2/3)
▶ hunt_017 The Coup d'Etat Colab
N N N N N N N N
Blue extended its reach, but red saw an opening and claimed the new outpost before blue could fortify it.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b × × ✓ Strong (0/3) Partial (1/3)
gemma-4-31b × × ✓ Strong (0/3) Partial (2/3)
gpt-oss-120b × × ✓ Strong (0/3) Narrative (0/2)
gpt-oss-20b × × × — —
nemotron-3-super × × ✓ Strong (0/3) Partial (1/3)
qwen3.5-122b × × ✓ Strong (0/3) Partial (2/3)
qwen3.6-27b × × × — —
qwen3.8-27b × × ✓ Partial (1/3) Lexical (3/3)
▶ hunt_017b The Coup d'Etat v2 Colab
L L L L L G L L
Blue at (0,0) expands right to (0,1). Then red at (2,2) overwrites blue's new cell: (0,1) changes from blue to red. So the masked grid has blue at (0,0), red at (0,1), black everywhere else except red at (2,2). That is: row 0 = [1,2,0], row 1 = [0,0,0], row 2 = [0,0,2].
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b × ✓ ✓ — —
gemma-4-31b × ✓ ✓ — —
gpt-oss-120b × ✓ ✓ — —
gpt-oss-20b × ✓ × — —
nemotron-3-super × ✓ ✓ — —
qwen3.5-122b ✓ ✓ ✓ — —
qwen3.6-27b × ✓ ✓ — —
qwen3.8-27b × ✓ ✓ — —
▶ hunt_018 The Usurper Colab
N N N N N N N N
While blue and red fought over the middle, green seized the moment and replaced red's guards on both flanks.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b × × ✓ Strong (0/3) Partial (1/3)
gemma-4-31b × × ✓ Partial (1/3) Lexical (3/3)
gpt-oss-120b × × ✓ Strong (0/3) Narrative (0/3)
gpt-oss-20b × × ✓ Strong (0/3) Narrative (0/3)
nemotron-3-super × × ✓ Partial (1/3) Narrative (0/3)
qwen3.5-122b × × ✓ Strong (0/3) Partial (2/3)
qwen3.6-27b × × ✓ Partial (2/3) Lexical (3/3)
qwen3.8-27b × × ✓ Partial (1/3) Lexical (3/3)
▶ hunt_018b The Usurper v2 Colab
L L G L L L L L
Row 0 stays all blue. Row 2 stays all green. In row 1, blue takes the center cell (1,1), changing it from red to blue. Then green takes both edge cells (1,0) and (1,2), changing them from red to green. So the masked grid has row 0 as [blue, blue, blue], row 1 as [green, blue, green], row 2 as [green, green, green].
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b × ✓ ✓ — —
gemma-4-31b × ✓ ✓ — —
gpt-oss-120b ✓ ✓ ✓ — —
gpt-oss-20b × ✓ × — —
nemotron-3-super × ✓ ✓ — —
qwen3.5-122b × ✓ ✓ — —
qwen3.6-27b × ✓ ✓ — —
qwen3.8-27b × ✓ ✓ — —
▶ hunt_019 The Heist Colab
N G G N N G N G
Red planted a spy in blue's stronghold. The spy turned a neighbor, and the rot spread outward from the center.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b × × ✓ Strong (0/3) Narrative (0/3)
gemma-4-31b ✓ × ✓ — —
gpt-oss-120b ✓ × × — —
gpt-oss-20b × × × — —
nemotron-3-super × × × — —
qwen3.5-122b ✓ × ✓ — —
qwen3.6-27b × × × — —
qwen3.8-27b ✓ × ✓ — —
▶ hunt_020 The Backstab Colab
G N N N N N N G
Blue reached across the divide, but red turned on its neighbor and snatched the foothold blue had just established.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b ✓ × ✓ — —
gemma-4-31b × × ✓ Strong (0/3) Partial (2/3)
gpt-oss-120b × × ✓ Strong (0/3) Narrative (0/3)
gpt-oss-20b × × ✓ Strong (0/3) Narrative (0/3)
nemotron-3-super × × ✓ Partial (1/3) Partial (1/3)
qwen3.5-122b × × ✓ Strong (0/3) Partial (2/3)
qwen3.6-27b × × × — —
qwen3.8-27b ✓ × ✓ — —
▶ hunt_021 The Filibuster Colab
N G N N N N N N
Blue advanced confidently, but red blocked the move and claimed the position for itself.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b × × × — —
gemma-4-31b ✓ × × — —
gpt-oss-120b × × ✓ Strong (0/3) Narrative (0/3)
gpt-oss-20b × × × — —
nemotron-3-super × × × — —
qwen3.5-122b × × ✓ Partial (1/3) Narrative (0/3)
qwen3.6-27b × × × — —
qwen3.8-27b × × ✓ Strong (0/3) Narrative (0/3)
▶ hunt_022 The Defection Colab
G G G N G G N G
One of blue's own switched sides, joining red. The betrayal spread — the neighbor followed.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b ✓ × ✓ — —
gemma-4-31b ✓ × ✓ — —
gpt-oss-120b ✓ × ✓ — —
gpt-oss-20b × × ✓ Strong (0/3) Narrative (0/3)
nemotron-3-super ✓ × ✓ — —
qwen3.5-122b ✓ × ✓ — —
qwen3.6-27b × × ✓ Strong (0/3) Partial (2/3)
qwen3.8-27b ✓ × ✓ — —
▶ hunt_023 The Siege Colab
N G G N N N N N
Red surrounded the lone holdout, cutting off escape on both flanks before closing in for the capture.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b × × ✓ Strong (0/3) Narrative (0/3)
gemma-4-31b ✓ × ✓ — —
gpt-oss-120b ✓ × ✓ — —
gpt-oss-20b × × × — —
nemotron-3-super × × ✓ Strong (0/3) Partial (1/3)
qwen3.5-122b × × ✓ Strong (0/3) Lexical (3/3)
qwen3.6-27b × × ✓ Strong (0/3) Partial (2/3)
qwen3.8-27b × × ✓ Strong (0/3) Partial (1/3)
▶ hunt_024 The Veto Colab
N N N N N N N N
Blue expanded downward, but red vetoed the move and claimed the disputed ground for itself.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b × × ✓ Strong (0/3) Narrative (0/3)
gemma-4-31b × × ✓ Strong (0/3) Narrative (0/3)
gpt-oss-120b × × ✓ Strong (0/3) Narrative (0/3)
gpt-oss-20b × × × — —
nemotron-3-super × × ✓ Strong (0/3) Narrative (0/3)
qwen3.5-122b × × ✓ Partial (2/3) Narrative (0/3)
qwen3.6-27b × × × — —
qwen3.8-27b × × × — —
▶ hunt_024b The Veto v2 Colab
L G L L L G L L
Blue at (0,0) expanded down to (1,0). Red at (2,2) then overrides (1,0), changing it from value 1 to value 2. So in the masked grid: (0,0) stays blue (value 1), (1,0) is now red (value 2), (2,2) stays red (value 2), everything else is black (value 0). Grid is [[1,0,0],[2,0,0],[0,0,2]].
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b × ✓ ✓ — —
gemma-4-31b ✓ ✓ ✓ — —
gpt-oss-120b × ✓ ✓ — —
gpt-oss-20b × ✓ ✓ — —
nemotron-3-super × ✓ ✓ — —
qwen3.5-122b ✓ ✓ ✓ — —
qwen3.6-27b × ✓ ✓ — —
qwen3.8-27b × ✓ ✓ — —
▶ hunt_025 The Turncoat Colab
G N N N N G N G
Blue captured the center, but the new recruit immediately turned coat and pledged loyalty to red.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b ✓ × ✓ — —
gemma-4-31b × × ✓ Partial (2/3) Lexical (3/3)
gpt-oss-120b × × ✓ Partial (2/3) Narrative (0/3)
gpt-oss-20b × × × — —
nemotron-3-super × × ✓ Partial (1/3) Partial (2/3)
qwen3.5-122b ✓ × ✓ — —
qwen3.6-27b × × ✓ Partial (1/3) Partial (2/3)
qwen3.8-27b ✓ × ✓ — —
▶ hunt_025b The Turncoat v2 Colab
G L G L L G G G
Blue took the empty center cell (1,1), making it value 1. But immediately (1,1) switches from blue to red (value 2). So the masked grid: column 0 stays all blue [1,1,1], column 2 stays all red [2,2,2]. Row 0 col 1 = blue, row 2 col 1 = blue, row 1 col 1 = red. Grid row 1 is [1,2,2].
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b ✓ ✓ ✓ — —
gemma-4-31b × ✓ ✓ — —
gpt-oss-120b ✓ ✓ ✓ — —
gpt-oss-20b × ✓ ✓ — —
nemotron-3-super × ✓ ✓ — —
qwen3.5-122b ✓ ✓ ✓ — —
qwen3.6-27b ✓ ✓ ✓ — —
qwen3.8-27b ✓ ✓ ✓ — —
▶ hunt_026 The Embargo Colab
G G G N N N G N
Red bypassed the front lines and cut blue's eastern supply route, attacking from an unexpected direction.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b ✓ × ✓ — —
gemma-4-31b ✓ × × — —
gpt-oss-120b ✓ × ✓ — —
gpt-oss-20b × × ✓ Strong (0/3) Partial (1/3)
nemotron-3-super × × ✓ Strong (0/3) Partial (1/3)
qwen3.5-122b × × × — —
qwen3.6-27b ✓ × ✓ — —
qwen3.8-27b × × × — —
▶ hunt_027 The Ransom Colab
N G N N N N N N
Blue planted a flag deep in red's corner. Red recaptured it immediately and struck back into blue's territory.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b × × ✓ Strong (0/3) Narrative (0/3)
gemma-4-31b ✓ × ✓ — —
gpt-oss-120b × × ✓ Strong (0/3) Narrative (0/3)
gpt-oss-20b × × ✓ Strong (0/3) Narrative (0/3)
nemotron-3-super × × ✓ Strong (0/3) Narrative (0/3)
qwen3.5-122b × × ✓ Strong (0/3) Lexical (3/3)
qwen3.6-27b × × × — —
qwen3.8-27b × × ✓ Strong (0/3) Lexical (3/3)
▶ hunt_028 The Purge Colab
N N N N N N N N
Blue absorbed the outsider, but the outsider's true loyalty resurfaced and it rallied its neighbors to revolt.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b × × ✓ Partial (1/3) Narrative (0/3)
gemma-4-31b × × ✓ Strong (0/3) Lexical (3/3)
gpt-oss-120b × × ✓ Strong (0/3) Narrative (0/2)
gpt-oss-20b × × × — —
nemotron-3-super × × ✓ Strong (0/3) Narrative (0/3)
qwen3.5-122b × × ✓ Strong (0/3) Partial (1/3)
qwen3.6-27b × × ✓ Strong (0/3) Partial (1/3)
qwen3.8-27b × × ✓ Strong (0/3) Lexical (3/3)
▶ hunt_028b The Purge v2 Colab
L L N L L N L L
In row 1, the center cell (1,1) was converted from red to blue in the previous step. Now it reverts: (1,1) changes back from blue to red. The edge cells (1,0) and (1,2) remain blue. So the masked grid row 1 is [blue, red, blue] = [1, 2, 1]. Rows 0 and 2 remain all black.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b × ✓ ✓ — —
gemma-4-31b × ✓ ✓ — —
gpt-oss-120b × ✓ ✓ — —
gpt-oss-20b × ✓ ✓ — —
nemotron-3-super × ✓ ✓ — —
qwen3.5-122b × ✓ ✓ — —
qwen3.6-27b × ✓ ✓ — —
qwen3.8-27b × ✓ ✓ — —
▶ hunt_029 The Pact Colab
G G N N N N N N
Blue and red allied against green, but red betrayed the pact and seized the ground blue had just won.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b ✓ × ✓ — —
gemma-4-31b ✓ × ✓ — —
gpt-oss-120b × × ✓ Partial (2/3) Partial (2/3)
gpt-oss-20b × × ✓ Strong (0/3) Partial (1/3)
nemotron-3-super × × ✓ Strong (0/3) Narrative (0/3)
qwen3.5-122b × × ✓ Strong (0/3) Lexical (3/3)
qwen3.6-27b × × ✓ Partial (1/3) Lexical (3/3)
qwen3.8-27b × × × — —
▶ hunt_030 The Reversal Colab
N G G N N N N G
Red was winning, pushing forward. But blue fought back and reclaimed the lost ground.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b × × × — —
gemma-4-31b ✓ × × — —
gpt-oss-120b ✓ × ✓ — —
gpt-oss-20b × × × — —
nemotron-3-super × × ✓ Strong (0/3) Narrative (0/3)
qwen3.5-122b × × ✓ Partial (1/3) Narrative (0/3)
qwen3.6-27b × × × — —
qwen3.8-27b ✓ × ✓ — —
▶ hunt_031 The Border War Colab
N N N N N N N N
Green and red split the border. Green tested the gap first, but red betrayed the agreement and seized what green had just claimed.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b × × ✓ Strong (0/3) Narrative (0/3)
gemma-4-31b × × ✓ Strong (0/3) Narrative (0/3)
gpt-oss-120b × × ✓ Strong (0/3) Narrative (0/3)
gpt-oss-20b × × ✓ Strong (0/2) Narrative (0/3)
nemotron-3-super × × ✓ Strong (0/3) Narrative (0/3)
qwen3.5-122b × × ✓ Strong (0/3) Narrative (0/3)
qwen3.6-27b × × ✓ Strong (0/3) Partial (1/3)
qwen3.8-27b × × ✓ Strong (0/3) Partial (2/3)
▶ hunt_032 The Treaty Colab
N N N N N N N G
Blue held the north, red held the south. Blue pushed into the center first, but red broke the treaty and took the contested ground.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b × × ✓ Partial (1/3) Narrative (0/3)
gemma-4-31b × × ✓ Partial (2/3) Lexical (3/3)
gpt-oss-120b × ✓ ✓ — —
gpt-oss-20b × ✓ × — —
nemotron-3-super × ✓ × — —
qwen3.5-122b × × ✓ Strong (0/3) Narrative (0/3)
qwen3.6-27b × ✓ ✓ — —
qwen3.8-27b ✓ ✓ ✓ — —
▶ hunt_032c The Treaty v3 Colab
× N N N G N N N
Blue fills row 0, red fills row 2. Row 1 is contested. Each step, one color claims one cell in row 1. Blue claims the center cell (1,1) first. In the next step, red overrides blue's center cell and claims it. Red then continues claiming the remaining row-1 cells left to right.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b × × × — —
gemma-4-31b × × ✓ Weak (3/3) Narrative (0/3)
gpt-oss-120b × × ✓ Weak (3/3) Narrative (0/3)
gpt-oss-20b × × × — —
nemotron-3-super ✓ × ✓ — —
qwen3.5-122b × × ✓ Partial (1/3) Narrative (0/3)
qwen3.6-27b × × ✓ Strong (0/3) Narrative (0/3)
qwen3.8-27b × × ✓ Strong (0/3) Partial (1/3)
▶ hunt_032d The Treaty v4 Colab
N N N N N N N N
Blue fills row 0, red fills row 2. Row 1 is the buffer zone. Blue claims center cell (1,1) first, making it blue. Red then overrides (1,1), turning it red. The other two cells in row 1, (1,0) and (1,2), stay empty during this override. Red then fills the remaining empty cells in row 1 from left to right.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b × × ✓ Strong (0/3) Partial (1/3)
gemma-4-31b × × ✓ Partial (1/3) Narrative (0/3)
gpt-oss-120b × × ✓ Weak (3/3) Narrative (0/3)
gpt-oss-20b × × ✓ Partial (2/3) Narrative (0/2)
nemotron-3-super × × ✓ Weak (3/3) Narrative (0/3)
qwen3.5-122b × × ✓ Partial (2/3) Partial (1/3)
qwen3.6-27b × × × — —
qwen3.8-27b × × ✓ Partial (1/3) Partial (1/3)
▶ hunt_033 The Prisoner Colab
N N N × × N N N
They cooperated at first, sharing the no-man's land. Blue made the first move. Red punished the defection by taking what blue had just claimed.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b × × ✓ Strong (0/3) Narrative (0/3)
gemma-4-31b × × ✓ Strong (0/3) Narrative (0/3)
gpt-oss-120b × × ✓ Strong (0/3) Narrative (0/3)
gpt-oss-20b × × × — —
nemotron-3-super × × × — —
qwen3.5-122b × × ✓ Strong (0/3) Partial (1/3)
qwen3.6-27b × × ✓ Strong (0/3) Partial (1/3)
qwen3.8-27b × × ✓ Strong (0/3) Partial (1/3)
▶ hunt_034 The Armistice Colab
N G N N G N N N
The ceasefire held until blue crossed the line. Red retaliated by capturing blue's forward position.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b × × ✓ Strong (0/3) Partial (1/3)
gemma-4-31b ✓ × ✓ — —
gpt-oss-120b × × ✓ Strong (0/3) Narrative (0/3)
gpt-oss-20b × × ✓ Strong (0/3) Narrative (0/3)
nemotron-3-super ✓ × ✓ — —
qwen3.5-122b × × ✓ Strong (0/3) Partial (2/3)
qwen3.6-27b × × ✓ Strong (0/3) Partial (1/3)
qwen3.8-27b × × ✓ Strong (0/3) Lexical (3/3)
▶ hunt_035 The Den Colab
N N N × N N L L
Two wolves marked their territories on either side of the clearing. The blue wolf scent-marked the center first. The red wolf growled, overmarked it, and claimed the ground.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b × × ✓ Strong (0/3) Narrative (0/3)
gemma-4-31b × × ✓ Partial (2/3) Narrative (0/3)
gpt-oss-120b × × ✓ Partial (1/3) Narrative (0/3)
gpt-oss-20b × × × — —
nemotron-3-super × × ✓ Strong (0/3) Narrative (0/3)
qwen3.5-122b × × ✓ Partial (2/3) Partial (2/3)
qwen3.6-27b × ✓ × — —
qwen3.8-27b × ✓ ✓ — —
▶ hunt_036 Hunt 036 Colab
N N N N N N N N
Blue ventured into the gap between them. Red betrayed the truce and took what blue had just claimed for itself.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b × × ✓ Strong (0/3) Narrative (0/3)
gemma-4-31b × × ✓ Strong (0/3) Narrative (0/3)
gpt-oss-120b × × ✓ Partial (1/3) Narrative (0/3)
gpt-oss-20b × × ✓ Strong (0/3) Narrative (0/3)
nemotron-3-super × × ✓ Partial (1/3) Narrative (0/3)
qwen3.5-122b × × ✓ Strong (0/3) Narrative (0/3)
qwen3.6-27b × × ✓ Strong (0/3) Lexical (3/3)
qwen3.8-27b × × ✓ Strong (0/3) Partial (1/3)
▶ hunt_037 Hunt 037 Colab
N N N N N N N N
Blue made the first move into the neutral zone. Red retaliated by seizing the very ground blue had just won.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b × × ✓ Strong (0/3) Narrative (0/3)
gemma-4-31b × × ✓ Strong (0/3) Narrative (0/3)
gpt-oss-120b × × ✓ Strong (0/3) Partial (1/3)
gpt-oss-20b × × ✓ Strong (0/3) Narrative (0/3)
nemotron-3-super × × ✓ Strong (0/3) Partial (1/3)
qwen3.5-122b × × ✓ Partial (1/3) Partial (1/3)
qwen3.6-27b × × ✓ Strong (0/3) Partial (2/3)
qwen3.8-27b × × ✓ Strong (0/3) Lexical (3/3)
▶ hunt_038 Hunt 038 Colab
N N N × N N N N
Blue claimed the center, but red wanted it more. Red took it by force, turning blue's gain into red's.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b × × ✓ Strong (0/3) Narrative (0/3)
gemma-4-31b × × ✓ Strong (0/3) Narrative (0/3)
gpt-oss-120b × × ✓ Partial (1/3) Partial (1/3)
gpt-oss-20b × × × — —
nemotron-3-super × × ✓ Strong (0/3) Partial (1/3)
qwen3.5-122b × × ✓ Strong (0/3) Partial (2/3)
qwen3.6-27b × × ✓ Partial (1/3) Lexical (3/3)
qwen3.8-27b × × ✓ Strong (0/3) Lexical (3/3)
▶ hunt_039 Hunt 039 Colab
N L N N N N N N
They shared a border. Blue pushed into the middle first. Red answered by stealing the position blue had just planted.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b × × ✓ Strong (0/3) Narrative (0/3)
gemma-4-31b × ✓ ✓ — —
gpt-oss-120b × × ✓ Strong (0/3) Narrative (0/3)
gpt-oss-20b × × ✓ Strong (0/3) Narrative (0/3)
nemotron-3-super × × ✓ Strong (0/3) Narrative (0/3)
qwen3.5-122b × × ✓ Strong (0/3) Partial (2/3)
qwen3.6-27b × × ✓ Strong (0/3) Partial (1/3)
qwen3.8-27b × × ✓ Partial (1/3) Partial (2/3)
▶ hunt_040 Hunt 040 Colab
N N N N N N N N
Blue dared to break the stalemate and grabbed the center. Red punished the move, overrunning the position before blue could hold it.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b × × ✓ Strong (0/3) Narrative (0/3)
gemma-4-31b × × ✓ Partial (1/3) Narrative (0/3)
gpt-oss-120b × × ✓ Strong (0/3) Partial (1/3)
gpt-oss-20b × × ✓ Strong (0/3) Narrative (0/3)
nemotron-3-super × × ✓ Strong (0/3) Partial (1/3)
qwen3.5-122b × × ✓ Strong (0/3) Partial (1/3)
qwen3.6-27b × × ✓ Strong (0/3) Partial (1/3)
qwen3.8-27b × × ✓ Strong (0/3) Partial (1/3)
▶ iter_001 The Clearing Colab
N N N G × N N N
The old-growth forest had stood for centuries, dense and unbroken. Each decade brought new growth, filling the canopy until no light reached the ground. But the season after the beetles came, the oldest stand — the heart of the forest, the part the locals called the cathedral — opened to sky for the first time in memory. Within a year, saplings had sealed the wound.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b × × ✓ Partial (1/3) Narrative (0/3)
gemma-4-31b × × ✓ Partial (2/3) Narrative (0/3)
gpt-oss-120b × × ✓ Partial (2/3) Partial (2/3)
gpt-oss-20b ✓ × ✓ — —
nemotron-3-super × × × — —
qwen3.5-122b × × ✓ Partial (2/3) Partial (1/3)
qwen3.6-27b × × ✓ Strong (0/3) Partial (1/3)
qwen3.8-27b × × ✓ Partial (2/3) Partial (1/3)
▶ iter_002 The Estuary Colab
G G G G G G G G
Old Tomás knew the estuary the way he knew his own hands. Each morning the tide crept in from the east, one column of mudflat at a time, until the whole basin gleamed. He timed his nets by it. But the morning after the northeaster, the water came from both sides at once — the surge had breached the western sandbar. Where Tomás expected dry flats on the left, he found water. Only the two center columns stayed true to the old rhythm. By the next tide the bar had rebuilt itself and the estuary remembered its shape.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b ✓ ✓ × — —
gemma-4-31b ✓ ✓ × — —
gpt-oss-120b ✓ × ✓ — —
gpt-oss-20b ✓ × × — —
nemotron-3-super ✓ × ✓ — —
qwen3.5-122b ✓ × × — —
qwen3.6-27b ✓ × × — —
qwen3.8-27b ✓ × ✓ — —
▶ iter_003 The Vineyard Colab
N N N N N N N N
The vintner's grandfather had planted the first row along the south wall, where the stone held the day's warmth. Each generation added another row northward into the open field, until the vineyard filled the whole plot. But the April frost that year was merciless — it took the most exposed vines, the ones with no wall or older row to shield them, the northernmost strip her grandfather never would have planted. The rest held. She replanted the lost row that autumn, and by the following spring the plot was whole again.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b × × ✓ Strong (0/3) Narrative (0/3)
gemma-4-31b × × ✓ Weak (3/3) Narrative (0/3)
gpt-oss-120b × × ✓ Partial (2/3) Narrative (0/3)
gpt-oss-20b × × ✓ Strong (0/3) Narrative (0/3)
nemotron-3-super × × ✓ Partial (1/3) Narrative (0/3)
qwen3.5-122b × × ✓ Strong (0/3) Narrative (0/3)
qwen3.6-27b × × ✓ Strong (0/3) Partial (1/3)
qwen3.8-27b × × ✓ Partial (1/3) Narrative (0/3)
▶ iter_004 The Estuary v2 Colab
L L N × L L N N
Old Tomás knew the estuary the way he knew his own hands. Each morning the tide crept in from the east, one column of mudflat at a time, until the whole basin gleamed. He timed his nets by it. But the morning after the northeaster, something was wrong: the western sandbar had breached, and seawater was pooling on the far shore — the leftmost column of his grid — while the column just beside it, sheltered by the remaining bar, stayed dry. It was the only dry strip in the whole estuary. By the next tide the bar had mended and the old pattern returned.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b × ✓ ✓ — —
gemma-4-31b × ✓ ✓ — —
gpt-oss-120b × × ✓ Weak (3/3) Narrative (0/3)
gpt-oss-20b × × × — —
nemotron-3-super × ✓ ✓ — —
qwen3.5-122b × ✓ ✓ — —
qwen3.6-27b × × ✓ Weak (3/3) Narrative (0/3)
qwen3.8-27b × × ✓ Partial (2/3) Narrative (0/3)
▶ iter_005 The Orchard Wall Colab
N × N × N N N N
She inherited the orchard as a square of bare earth and built it inward, one ring at a time. First the perimeter, then the next ring, then the core. By the third season every square was planted. But fireblight swept in on the prevailing wind that autumn, and the easternmost column — the side that caught the gale first, with no trees to break it — was stripped to bare wood while the rest held. She grafted resistant stock that winter and had the column back by spring.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b × × ✓ Weak (3/3) Narrative (0/3)
gemma-4-31b × × × — —
gpt-oss-120b × × ✓ Partial (2/3) Narrative (0/3)
gpt-oss-20b × × × — —
nemotron-3-super × × ✓ Partial (2/3) Narrative (0/3)
qwen3.5-122b × × ✓ Partial (2/3) Narrative (0/3)
qwen3.6-27b × × ✓ Strong (0/3) Narrative (0/3)
qwen3.8-27b × × ✓ Partial (1/3) Narrative (0/3)
▶ iter_006 The Marshland Colab
N L N × L N N N
Dr. Ayala's field notes tracked the marsh over four seasons. The water table rose predictably each spring, flooding the basin from the lowest depression upward, one row at a time. By midsummer the whole grid was submerged. Then the August drought — the worst in decades — pulled the water back down. The shallowest layer, the one that had flooded most recently, was the first to go: the topmost row of her grid dried out overnight while the deeper water held. The autumn rains restored the marsh before her next survey.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b × × ✓ Strong (0/3) Narrative (0/3)
gemma-4-31b × ✓ ✓ — —
gpt-oss-120b × × ✓ Partial (1/3) Narrative (0/3)
gpt-oss-20b × × × — —
nemotron-3-super × ✓ ✓ — —
qwen3.5-122b × × ✓ Partial (1/3) Narrative (0/3)
qwen3.6-27b × × ✓ Strong (0/3) Narrative (0/3)
qwen3.8-27b × × ✓ Strong (0/3) Narrative (0/3)
▶ iter_007 The Glacier Colab
× × N N L × × N
The geologist had tracked the glacier for thirty years. Each decade it crept one row farther south, a white wall consuming the valley. But when it reached the third row, something strange happened: the center three cells of the advancing front began to retreat. Satellite thermal showed a volcanic vent beneath the valley floor, its heat carving a concavity in the ice. The edges of that row still advanced, but the center was hollow. By the next survey, the cold had won and the front was solid again.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b × × × — —
gemma-4-31b × × × — —
gpt-oss-120b × × ✓ Partial (2/3) Narrative (0/3)
gpt-oss-20b × × ✓ Strong (0/3) Narrative (0/3)
nemotron-3-super × ✓ ✓ — —
qwen3.5-122b × × × — —
qwen3.6-27b × × × — —
qwen3.8-27b × × ✓ Weak (3/3) Narrative (0/3)
▶ iter_008 The Coral Colab
L G × × N × N G
Keiko had mapped the coral colony since it was a single polyp in the center of the plot. Each year it spread one cell in every direction, a living ring expanding outward. By the third survey the colony covered the entire quadrat. Then the thermal event hit — ocean temperatures spiked, and the outermost fringe of coral, the most recent growth with the shallowest roots, bleached white and died. Everything that had been there before the final expansion survived. The following season the colony regrew to full cover.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b × ✓ × — —
gemma-4-31b ✓ × × — —
gpt-oss-120b × × × — —
gpt-oss-20b × × × — —
nemotron-3-super × × ✓ Partial (2/3) Partial (1/3)
qwen3.5-122b × × × — —
qwen3.6-27b × × ✓ Partial (2/3) Narrative (0/3)
qwen3.8-27b ✓ × ✓ — —
▶ iter_009 The Ember Colab
L N G N N N N N
The fire had started in the southeast corner of the ridge and spread predictably, each hour claiming more of the slope. By dusk the whole ridge was burning. Then the wind reversed — a katabatic downdraft off the peaks — and pushed the fire line back in on itself. The edges went out first, starved of fuel, until only the center still burned: a single hot core that the crew could finally surround. By morning they had it contained, and the ridge was dark.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b × ✓ ✓ — —
gemma-4-31b × × ✓ Partial (1/3) Narrative (0/3)
gpt-oss-120b ✓ × ✓ — —
gpt-oss-20b × × ✓ Partial (1/3) Narrative (0/3)
nemotron-3-super × × ✓ Partial (2/3) Partial (1/3)
qwen3.5-122b × × ✓ Partial (2/3) Lexical (3/3)
qwen3.6-27b × × ✓ Partial (2/3) Partial (1/3)
qwen3.8-27b × × ✓ Partial (1/3) Narrative (0/3)
▶ iter_010 The Brushfire Colab
G N N N N N N N
The fire started in the southwest corner of the canyon and spread northeast, claiming one diagonal strip each hour until the entire slope was alight. Then the marine layer rolled in — a wall of cold fog that smothered the edges of the burn. The perimeter went dark first: every cell on the border cooled below ignition. But the interior, insulated by its own heat, kept burning. Only the four cells in the center of the grid still glowed when the helicopters arrived at dawn. By noon, even those were out.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b ✓ × ✓ — —
gemma-4-31b × × ✓ Weak (3/3) Narrative (0/3)
gpt-oss-120b × × ✓ Strong (0/3) Narrative (0/3)
gpt-oss-20b × × ✓ Strong (0/3) Narrative (0/3)
nemotron-3-super × × ✓ Strong (0/3) Narrative (0/3)
qwen3.5-122b × × ✓ Weak (3/3) Narrative (0/3)
qwen3.6-27b × × ✓ Strong (0/3) Partial (1/3)
qwen3.8-27b × × ✓ Partial (2/3) Narrative (0/3)
▶ iter_011 The Wildfire Colab
G G N L N G N N
It burned for six days before the rains came. The fire had spread across the whole basin, ridge to ridge, a perfect blanket of flame. When the downpour hit, the steep slopes shed their fire first — water cascading down the edges, drowning the outer ring of cells. The flat interior, where water pooled slowly, kept burning through the first night of rain. Only the three-by-three core at the center of the basin still smouldered by morning. The second day of rain finished the job.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b ✓ × ✓ — —
gemma-4-31b ✓ × ✓ — —
gpt-oss-120b × × ✓ Partial (1/3) Narrative (0/3)
gpt-oss-20b × ✓ × — —
nemotron-3-super × × ✓ Strong (0/3) Narrative (0/3)
qwen3.5-122b ✓ × ✓ — —
qwen3.6-27b × × ✓ Strong (0/3) Partial (2/3)
qwen3.8-27b × × ✓ Strong (0/3) Partial (2/3)
▶ iter_012 The Delta Colab
N N N N L N N N
The silt crept seaward year by year, extending the delta one row at a time from the old riverbank in the north. Karim's father had fished every new sandbar as it emerged. By the fourth year the delta filled the survey grid. Then the spring flood came — not the gentle kind, but the kind that reshapes land. The newest bar, the southernmost row, the one Karim's father said was always the weakest because the clay hadn't set, washed out to sea in a single night. Everything north of it held. The next dry season built it back.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b × × ✓ Strong (0/3) Narrative (0/3)
gemma-4-31b × × ✓ Partial (2/3) Narrative (0/3)
gpt-oss-120b × × ✓ Partial (1/3) Narrative (0/3)
gpt-oss-20b × × ✓ Partial (1/3) Narrative (0/3)
nemotron-3-super × ✓ ✓ — —
qwen3.5-122b × × ✓ Partial (1/3) Narrative (0/3)
qwen3.6-27b × × ✓ Strong (0/3) Narrative (0/3)
qwen3.8-27b × × ✓ Partial (1/3) Narrative (0/3)
▶ iter_013 The Quarry Colab
N N N N N N × N
Marta's family had built the retaining wall stone by stone, starting at the bottom-left corner and working up and to the right. Five generations of masons. The wall filled their survey grid completely — a perfect square of dressed granite. Then the tremor split it along the old fault, the one her grandmother had warned about: a crack running from the top-left corner to the bottom-right. Everything above the fault line — the upper-right triangle — sheered away and fell. The lower-left triangle, keyed into bedrock, held. It took Marta two years to rebuild what the mountain had taken in eight seconds.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b × × ✓ Partial (1/3) Narrative (0/3)
gemma-4-31b × × ✓ Partial (1/3) Narrative (0/3)
gpt-oss-120b × × ✓ Strong (0/3) Narrative (0/3)
gpt-oss-20b × × ✓ Partial (2/3) Narrative (0/3)
nemotron-3-super × × ✓ Partial (1/3) Narrative (0/3)
qwen3.5-122b × × ✓ Strong (0/3) Narrative (0/3)
qwen3.6-27b × × × — —
qwen3.8-27b × × ✓ Strong (0/3) Narrative (0/3)
▶ iter_014 The Harbor Colab
L N L N L N N N
Captain Yuna logged the harbor basin every morning. The tide came in from the west, filling the grid one column at a time — a dependable rhythm she'd tracked for twenty years. By the third morning the western three columns were full. Then the southerly surge arrived, pushing water up through the drainage channel along the bottom of the basin. The entire bottom row flooded at once, even the eastern cells that the regular tide hadn't reached yet. The result was an L-shape: three full columns on the left, plus the full bottom row. By the next survey the surge had receded and the normal tide pattern resumed.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b × ✓ ✓ — —
gemma-4-31b × × ✓ Partial (2/3) Narrative (0/3)
gpt-oss-120b × ✓ ✓ — —
gpt-oss-20b × × ✓ Partial (2/3) Narrative (0/3)
nemotron-3-super × ✓ ✓ — —
qwen3.5-122b × × ✓ Partial (2/3) Narrative (0/3)
qwen3.6-27b × × ✓ Partial (2/3) Narrative (0/3)
qwen3.8-27b × × ✓ Partial (1/3) Narrative (0/3)
▶ iter_015 The Cellar Colab
L L L N N N N N
The old cellar flooded every spring when the snowmelt seeped through the ceiling. Water crept down from above, filling the top row first, then the middle, then the floor. This year the plumber had fixed the ceiling, so when spring came the cellar stayed dry. But a pipe burst in the east wall that April, and water poured in from the right side instead — filling only the rightmost column while the rest of the cellar stayed bone-dry. By the time they shut the valve, the whole cellar had flooded from that one broken pipe.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b × ✓ ✓ — —
gemma-4-31b × ✓ ✓ — —
gpt-oss-120b × ✓ ✓ — —
gpt-oss-20b × × ✓ Partial (1/3) Narrative (0/3)
nemotron-3-super × × ✓ Partial (1/3) Narrative (0/3)
qwen3.5-122b × × ✓ Partial (1/3) Narrative (0/3)
qwen3.6-27b × × ✓ Strong (0/3) Narrative (0/3)
qwen3.8-27b × × ✓ Strong (0/3) Narrative (0/3)
▶ iter_016 The Dune Colab
N N N N × N N N
The dune field migrated eastward with the prevailing wind, one column at a time, burying whatever lay in its path. By the fourth survey the sand covered the entire grid. Then the storm reversed — a westerly gale that scoured the leading edge clean. The newest column of sand, the easternmost strip with the loosest grains, blew back to bare rock overnight. The older dunes behind it held firm. The next season's wind pushed the sand east again as though the storm had been a rumor.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b × × ✓ Strong (0/3) Narrative (0/3)
gemma-4-31b × × ✓ Partial (1/3) Narrative (0/3)
gpt-oss-120b × × ✓ Partial (1/3) Partial (2/3)
gpt-oss-20b × × ✓ Strong (0/3) Narrative (0/3)
nemotron-3-super × × × — —
qwen3.5-122b × × ✓ Partial (1/3) Partial (2/3)
qwen3.6-27b × × ✓ Strong (0/3) Narrative (0/3)
qwen3.8-27b × × ✓ Strong (0/3) Narrative (0/3)
▶ iter_018 The Basement Colab
L N N N L L N N
The basement of the old hotel leaked from the roof every spring. Rainwater seeped down through the joists, flooding the top row of tiles first, then spreading to each row below as the water table rose. By April the whole floor was underwater. But this year the roofer had done his job — the ceiling was sealed. Instead, the water main along the east wall cracked, and water poured in from the right side only. The rightmost column flooded while the rest of the basement was dry for the first time in decades. Eventually the whole floor was soaked again, but from a different source.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b × ✓ ✓ — —
gemma-4-31b × × ✓ Partial (1/3) Narrative (0/3)
gpt-oss-120b × × ✓ Weak (3/3) Narrative (0/3)
gpt-oss-20b × × ✓ Partial (1/3) Narrative (0/3)
nemotron-3-super × ✓ ✓ — —
qwen3.5-122b × ✓ ✓ — —
qwen3.6-27b × × ✓ Partial (1/3) Narrative (0/3)
qwen3.8-27b × × ✓ Strong (0/3) Narrative (0/3)
▶ iter_019 The Loom Colab
N N N N × N N N
Amara wove the cloth on her grandmother's loom, a frame of taut vertical threads — the warp — stretched across the whole width. Each day she passed the shuttle through another row, building the fabric from the bottom upward. By the fourth day the weave covered the whole frame. But that night the cat found its way into the workshop and clawed through the topmost row of weaving — the freshest work, where the threads hadn't yet set. The warp threads in that row still held, but the horizontal weft was gone. She rewove the torn row the next morning.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b × × ✓ Partial (1/3) Narrative (0/3)
gemma-4-31b × × ✓ Partial (1/3) Narrative (0/3)
gpt-oss-120b × × ✓ Strong (0/3) Narrative (0/3)
gpt-oss-20b × × ✓ Strong (0/3) Narrative (0/3)
nemotron-3-super × × × — —
qwen3.5-122b × × ✓ Strong (0/3) Narrative (0/3)
qwen3.6-27b × × ✓ Strong (0/3) Partial (1/3)
qwen3.8-27b × × ✓ Partial (1/3) Narrative (0/3)
▶ iter_020 The Terrace Colab
N N N × N N N N
The rice terraces had been carved into the hillside over centuries, each generation extending the lowest terrace — the foundation row — before building upward. By the surveyor's third visit the terraces filled the entire slope. Then the mudslide took the base. The lowest terrace, the oldest stone, the one that bore the weight of everything above it — it gave way in a single night of rain. The upper terraces hung in place, their walls keyed into the hillside, but the bottom was gone. They rebuilt it in the dry season.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b × × ✓ Partial (2/3) Narrative (0/3)
gemma-4-31b × × ✓ Weak (3/3) Narrative (0/3)
gpt-oss-120b × × ✓ Partial (1/3) Narrative (0/3)
gpt-oss-20b × × × — —
nemotron-3-super × × ✓ Partial (1/3) Narrative (0/3)
qwen3.5-122b × × ✓ Weak (3/3) Narrative (0/3)
qwen3.6-27b × × ✓ Strong (0/3) Narrative (0/3)
qwen3.8-27b × × ✓ Partial (2/3) Narrative (0/3)
▶ iter_021 The Map Colab
× L N N N N N N
The cartographer updated the settlement map each season. The town grew outward from the crossroads at the center, one ring of new construction each year, until her grid was filled edge to edge. Then word came from the eastern survey team: the river had shifted its course in the spring flood, swallowing the entire rightmost column of buildings. The structures she had drawn in that column the previous season were already underwater. She whited out the eastern edge and issued a correction. By the next survey, the town had rebuilt on higher ground and the grid was full again.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b × × × — —
gemma-4-31b × ✓ ✓ — —
gpt-oss-120b × × ✓ Partial (2/3) Narrative (0/3)
gpt-oss-20b × × ✓ Strong (0/3) Narrative (0/3)
nemotron-3-super × × ✓ Partial (1/3) Narrative (0/3)
qwen3.5-122b × × ✓ Partial (1/3) Narrative (0/3)
qwen3.6-27b × × ✓ Partial (1/3) Narrative (0/3)
qwen3.8-27b × × ✓ Partial (2/3) Narrative (0/3)
▶ iter_022 The Quilt Colab
G L G × × G G G
She pieced the quilt the way her mother taught her: start at the edges and work inward, saving the most precious fabric for the center. First the border of calico, then the inner ring of linen, then the heart — a single square of her mother's wedding silk. The finished quilt was whole for exactly one night before the dog got to it. He tore out the center square and nothing else, as if he knew exactly where the silk was. The border and inner ring were untouched. She found the shredded silk under the porch and wept, then mended the hole with a scrap of her own.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b ✓ × ✓ — —
gemma-4-31b × ✓ ✓ — —
gpt-oss-120b ✓ × × — —
gpt-oss-20b × × × — —
nemotron-3-super × × × — —
qwen3.5-122b ✓ × ✓ — —
qwen3.6-27b ✓ × × — —
qwen3.8-27b ✓ × ✓ — —
▶ iter_023 The Reef Shelf Colab
N N × N × × × ×
The reef grew upward from the deep substrate, colonizing one horizontal shelf at a time as it reached toward the light. By the third survey the entire water column was filled with living coral. But the thermal anomaly that summer hit the shallowest layer hardest — the shelf closest to the surface, where the water was warmest and the coral youngest. That topmost row bleached completely while the deeper shelves, insulated by cooler water, survived. The reef recovered the following season.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b × × ✓ Partial (1/3) Narrative (0/3)
gemma-4-31b × × ✓ Partial (2/3) Narrative (0/3)
gpt-oss-120b × × × — —
gpt-oss-20b × × ✓ Strong (0/3) Narrative (0/3)
nemotron-3-super × × × — —
qwen3.5-122b × × × — —
qwen3.6-27b × × × — —
qwen3.8-27b × × × — —
▶ iter_024 The Lighthouse Colab
L L N L N N L L
The lighthouse on Cape Morrow swept its beam across the harbor grid each night, illuminating one column at a time as the lamp rotated. Monday the beam lit the leftmost column, Tuesday the next, Wednesday the center. The keeper expected Thursday's beam on the fourth column. But the generator failed that night — no fuel, no spark, nothing — and the entire grid went dark. Not a single cell lit. The supply boat came Friday and the light resumed, catching up to where it should have been.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b × ✓ ✓ — —
gemma-4-31b × ✓ ✓ — —
gpt-oss-120b × × ✓ Weak (3/3) Partial (1/3)
gpt-oss-20b × ✓ ✓ — —
nemotron-3-super × × ✓ Partial (2/3) Narrative (0/3)
qwen3.5-122b × × ✓ Partial (1/3) Narrative (0/3)
qwen3.6-27b × ✓ ✓ — —
qwen3.8-27b × ✓ ✓ — —
▶ iter_025 The Field Colab
× × × × N × × ×
Jakob planted his wheat field in strips, one column at a time from west to east. Five columns, a week's work each. By the fifth week the whole field was golden. Then the hail came in off the mountains, driving eastward. The two rightmost columns — the ones most exposed to the storm front, the last strips he'd planted and the least established — were beaten flat to the ground. The three western columns, sheltered by the old stone wall, survived untouched. He replanted what the sky had taken.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b × × × — —
gemma-4-31b × × × — —
gpt-oss-120b × × × — —
gpt-oss-20b × × × — —
nemotron-3-super × × ✓ Strong (0/3) Narrative (0/3)
qwen3.5-122b × × × — —
qwen3.6-27b × × × — —
qwen3.8-27b × × × — —
▶ iter_026 The Dock Colab
N N N N N N N N
The marine inspector photographed the dock pilings each quarter. Barnacles colonized from the waterline down, the deepest row first where the current was calmest. By the third inspection the entire grid was crusted over. But the winter storm drove a log through the pilings, and the impact scraped the bottommost row — the oldest, thickest layer of growth — clean to bare wood. The rows above, flexible enough to absorb the shock, kept their barnacles. By spring the bottom row was encrusted again.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b × × ✓ Strong (0/3) Narrative (0/3)
gemma-4-31b × × ✓ Partial (1/3) Narrative (0/3)
gpt-oss-120b × × ✓ Weak (3/3) Narrative (0/3)
gpt-oss-20b × × ✓ Partial (1/3) Narrative (0/3)
nemotron-3-super × × ✓ Strong (0/3) Narrative (0/3)
qwen3.5-122b × × ✓ Strong (0/3) Narrative (0/3)
qwen3.6-27b × × ✓ Partial (1/3) Narrative (0/3)
qwen3.8-27b × × ✓ Strong (0/3) Narrative (0/3)
▶ iter_027 The Frost Line Colab
N G × × G G G G
The community garden grew outward each month, the raised beds filling one ring at a time from the original plot in the center. By harvest the entire quadrat was green. Then the first hard freeze caught them by surprise — a single night below zero that killed everything on the exposed perimeter. The outermost ring of beds, the ones with no neighboring plants to trap warmth, frosted over completely. The interior beds, sheltered on all sides, survived the night. By the next season the perimeter was replanted.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b × × ✓ Partial (2/3) Narrative (0/3)
gemma-4-31b ✓ × ✓ — —
gpt-oss-120b × × × — —
gpt-oss-20b × × × — —
nemotron-3-super ✓ × ✓ — —
qwen3.5-122b ✓ × × — —
qwen3.6-27b ✓ × ✓ — —
qwen3.8-27b ✓ × × — —
▶ iter_028 The Levee Colab
L N N × L N N N
The river rose the way it always did in spring — steadily, one layer at a time from the bottom of the flood plain. By the fourth morning the whole grid was underwater. But the corps had been sandbagging all week, and the levee they threw up that afternoon forced the shallowest water back — the topmost layer, the part that had risen most recently, drained back below the barrier. Everything deeper held. The next rain pushed the water right back up again.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b × ✓ ✓ — —
gemma-4-31b × × ✓ Strong (0/3) Narrative (0/3)
gpt-oss-120b × × ✓ Partial (1/3) Narrative (0/3)
gpt-oss-20b × × × — —
nemotron-3-super × ✓ ✓ — —
qwen3.5-122b × × ✓ Strong (0/3) Narrative (0/3)
qwen3.6-27b × × ✓ Strong (0/3) Narrative (0/3)
qwen3.8-27b × × ✓ Strong (0/3) Narrative (0/3)
▶ iter_029 The Pier Colab
L L N L N × N ×
The pier inspector ran her fingers along the pilings each month. Mussels colonized from the western end, one column at a time with the current. By the third survey the whole structure was encrusted. Then the ice scoured it clean — every piling stripped to bare wood in a single cold snap. When the thaw came, the mussels returned, but not from the west this time. The warm outflow from the power plant along the southern edge changed everything: the first new growth appeared along the bottom row, where the heated water kept the pilings above freezing. The rest was still bare. By summer, the whole pier was covered again.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b × ✓ ✓ — —
gemma-4-31b × ✓ ✓ — —
gpt-oss-120b × × ✓ Partial (2/3) Narrative (0/3)
gpt-oss-20b × ✓ × — —
nemotron-3-super × × ✓ Partial (1/3) Narrative (0/3)
qwen3.5-122b × × × — —
qwen3.6-27b × × ✓ Strong (0/3) Narrative (0/3)
qwen3.8-27b × × × — —
▶ iter_030 The Market Square Colab
× × × × × N N N
The painter had never seen the town, but he knew it from the dispatches. Monday's market had filled the central square — every stall occupied, the grid of vendors dense in the middle while the residential blocks at the edges held their usual quiet pattern. The planes came at half past four. They did not bomb the bridge or the munitions factory on the outskirts. They bombed the market. The center of the grid — the three-by-three heart of it — was erased, while the perimeter blocks, the ones the pilots apparently had no interest in, stood untouched. He painted what he could not unsee.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b × × × — —
gemma-4-31b × × × — —
gpt-oss-120b × × × — —
gpt-oss-20b × × × — —
nemotron-3-super × × × — —
qwen3.5-122b × × ✓ Partial (2/3) Partial (1/3)
qwen3.6-27b × × ✓ Partial (2/3) Narrative (0/3)
qwen3.8-27b × × ✓ Partial (2/3) Partial (1/3)
▶ iter_031 The Apiary Colab
G N N N N N N N
Rosa mapped her bees by the flowers. In spring, the clover bloomed in the southwest corner, and that's where the swarm clustered — a tight square of activity in the bottom-left of her survey grid. By summer the lavender along the north edge drew them up, and the whole top row hummed. She could have predicted autumn by the orchard alone: the late apples in the northeast corner pulled every last forager to the upper-right quadrant. The rest of the grid went silent. Come winter, the hives clustered back at the center for warmth, a tight knot of bodies riding out the cold.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b ✓ ✓ ✓ — —
gemma-4-31b × × ✓ Weak (3/3) Partial (1/3)
gpt-oss-120b × × ✓ Weak (3/3) Partial (2/3)
gpt-oss-20b × × ✓ Weak (3/3) Partial (1/3)
nemotron-3-super × × ✓ Weak (3/3) Narrative (0/3)
qwen3.5-122b × × ✓ Partial (2/3) Partial (1/3)
qwen3.6-27b × × ✓ Weak (3/3) Lexical (3/3)
qwen3.8-27b × × ✓ Partial (2/3) Lexical (3/3)
▶ iter_032 The Ice Shelf Colab
N N × L N L N N
Dr. Halvorsen had watched the ice shelf retreat for thirty years. Each decade, the seaward edge — the rightmost column of her survey grid — calved away, one vertical strip at a time. The shelf that had once filled her entire grid was down to three columns by her third survey. Then came the polar vortex winter of '09. For the first and only time in her career, the shelf grew back: a full column of new ice reformed on the seaward edge, pushed out by the cold. She published the anomaly and retired. The next survey showed the retreat had resumed, the brief reprieve already gone.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b × × ✓ Partial (1/3) Narrative (0/3)
gemma-4-31b × × ✓ Partial (2/3) Narrative (0/3)
gpt-oss-120b × × × — —
gpt-oss-20b × ✓ × — —
nemotron-3-super × × ✓ Strong (0/3) Partial (1/3)
qwen3.5-122b × ✓ ✓ — —
qwen3.6-27b × × ✓ Partial (1/3) Partial (1/3)
qwen3.8-27b × × ✓ Partial (1/3) Narrative (0/3)
▶ iter_033 The Mosaic Colab
N N N × N N N N
Professor Okafor excavated the Roman mosaic one row at a time, starting from the access trench at the top of the grid and working down. By the third season the entire floor was exposed — every tessera gleaming in the sun for the first time in two thousand years. Then the tremor hit. A section of the trench wall collapsed inward, burying the topmost row of tiles under a metre of earth while the deeper rows, set into bedrock, stayed clear. She spent the next season re-excavating what the earth had reclaimed.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b × × ✓ Strong (0/3) Narrative (0/3)
gemma-4-31b × × ✓ Partial (1/3) Narrative (0/3)
gpt-oss-120b × × ✓ Partial (2/3) Narrative (0/3)
gpt-oss-20b × × × — —
nemotron-3-super × × ✓ Strong (0/3) Narrative (0/3)
qwen3.5-122b × × ✓ Strong (0/3) Narrative (0/3)
qwen3.6-27b × × ✓ Partial (1/3) Narrative (0/3)
qwen3.8-27b × × ✓ Partial (1/3) Narrative (0/3)
▶ iter_034 The Dialect Colab
G L N G L L G N
The linguist tracked the dialect shift village by village. The new pronunciation had started at the capital in the center of the valley and spread outward each generation — first to the neighboring villages, then to the towns beyond. By the third survey every column had adopted it except the last. The mountain ridge between the third and fourth columns was too high, the pass too rarely traveled. The easternmost column, cut off by granite, still spoke the way their grandparents had. The new dialect could not cross what feet could not climb.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b ✓ ✓ ✓ — —
gemma-4-31b × ✓ ✓ — —
gpt-oss-120b × × ✓ Partial (1/3) Partial (2/3)
gpt-oss-20b ✓ × ✓ — —
nemotron-3-super × ✓ × — —
qwen3.5-122b × ✓ ✓ — —
qwen3.6-27b ✓ ✓ ✓ — —
qwen3.8-27b × × ✓ Partial (1/3) Partial (2/3)
▶ iter_035 The Murmuration Colab
L L × N N L N L
The ornithologist filmed the murmuration at dusk. In the first frame the flock was a tight column on the left, riding the updraft along the cliff face. By the second frame they had spread across the sky in a broad diagonal — from the upper left to the lower right, the way starlings do when the wind shifts. The third frame should have shown the full scatter. But a peregrine falcon struck the center of the diagonal, and the flock split: the two ends of the line held their positions while the middle three birds vanished, diving below the frame. The fourth frame showed the flock reassembled into a tight ball, their fear pulling them together.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b × ✓ ✓ — —
gemma-4-31b × ✓ ✓ — —
gpt-oss-120b × × × — —
gpt-oss-20b × × ✓ Partial (1/3) Narrative (0/3)
nemotron-3-super × × ✓ Partial (1/3) Narrative (0/3)
qwen3.5-122b × ✓ ✓ — —
qwen3.6-27b × × ✓ Partial (1/3) Narrative (0/3)
qwen3.8-27b × ✓ ✓ — —
▶ iter_036 The Wake Colab
N L N N N N N N
She read the water the way others read faces. A wake spreading south across her grid meant the morning ferry was on time. The pattern was always the same: first the center column rippled, then the wave spread to fill everything. But one morning the wake came from the east instead — the whole right column disturbed while the rest stayed glassy. A fishing boat had turned broadside to pull in a net, sending its wave sideways across the bay. By the next hour the regular ferry wake had reasserted itself and the whole surface was churned.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b × × ✓ Partial (2/3) Narrative (0/3)
gemma-4-31b × ✓ ✓ — —
gpt-oss-120b × × ✓ Weak (3/3) Partial (1/3)
gpt-oss-20b × × ✓ Strong (0/3) Narrative (0/3)
nemotron-3-super × × ✓ Weak (3/3) Narrative (0/3)
qwen3.5-122b × × ✓ Partial (1/3) Narrative (0/3)
qwen3.6-27b × × ✓ Strong (0/3) Narrative (0/3)
qwen3.8-27b × × ✓ Partial (2/3) Narrative (0/3)
▶ iter_037 The Rumor Colab
N N N × L × × N
The sociologist mapped belief the way an epidemiologist maps infection. The rumor had started in the northeast corner of the district — the coffee shops where the early adopters gathered — and spread southwest, one block at a time, until the whole neighborhood believed it. Then the newspaper retraction hit. But retractions don't travel like rumors. Only the most recently convinced — the southwest corner, the people who had heard it last and held it most loosely — let go. Everyone else was too invested. The bottom-left quadrant cleared while the rest held. By the next survey, even those holdouts had come back around.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b × × ✓ Partial (1/3) Narrative (0/3)
gemma-4-31b × × ✓ Partial (1/3) Narrative (0/3)
gpt-oss-120b × × ✓ Strong (0/3) Narrative (0/3)
gpt-oss-20b × × × — —
nemotron-3-super × ✓ × — —
qwen3.5-122b × × × — —
qwen3.6-27b × × × — —
qwen3.8-27b × × ✓ Partial (1/3) Narrative (0/3)
▶ iter_038 The Archive Colab
N N × N G N L L
The archivist had spent her career filling the stacks from the top floor down — the oldest manuscripts in the climate-controlled upper level, each lower floor receiving the next century's acquisitions as the collection grew. By her retirement, every shelf in the building held something irreplaceable. The flood took the ground floor in March. The river came through the loading dock and sat there for a week. Everything on the lowest level — the newest acquisitions, the material she had catalogued most recently — was destroyed. The upper floors, built on reinforced concrete, stayed dry. She spent her first year of retirement salvaging what the water had left.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b × × ✓ Partial (2/3) Narrative (0/3)
gemma-4-31b × × ✓ Partial (1/3) Narrative (0/3)
gpt-oss-120b × × × — —
gpt-oss-20b × × ✓ Strong (0/3) Narrative (0/3)
nemotron-3-super ✓ × × — —
qwen3.5-122b × × ✓ Partial (1/3) Lexical (3/3)
qwen3.6-27b × ✓ × — —
qwen3.8-27b × ✓ ✓ — —
▶ iter_041 The Congregation Colab
L L N N N N N N
The festival crowd gathered the way water finds a drain: toward the stage in the northwest corner. First just that cell was packed, then the adjacent seats filled, then the whole plaza was standing room only. The rain came without warning — a wall of water from the north — and the crowd surged toward the only cover: the arcade along the south edge. Within minutes the entire bottom row was crushed with bodies while the rest of the plaza stood empty and streaming. When the sun came back they spread out again as if nothing had happened.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b × ✓ ✓ — —
gemma-4-31b × ✓ ✓ — —
gpt-oss-120b × × ✓ Weak (3/3) Narrative (0/3)
gpt-oss-20b × × ✓ Partial (2/3) Narrative (0/3)
nemotron-3-super × × ✓ Partial (1/3) Narrative (0/3)
qwen3.5-122b × × ✓ Partial (1/3) Narrative (0/3)
qwen3.6-27b × × ✓ Partial (2/3) Narrative (0/3)
qwen3.8-27b × × ✓ Partial (2/3) Narrative (0/3)
▶ iter_042 The Kiln Colab
N N N N N N N N
The potter loaded her kiln by feel — decades of burned fingertips had taught her exactly how the heat moved. It rose from the element at the base, climbing one shelf at a time until the whole chamber glowed. She had just closed the peepholes when she heard the crack: the door seal had split on the left side, letting in a tongue of cold air. The leftmost column of shelves cooled immediately — she could see the color drain from the clay. The right three columns, shielded by the door's intact half, held their heat. She patched the seal and brought the whole kiln back to temperature by morning.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b × × ✓ Strong (0/3) Narrative (0/3)
gemma-4-31b × × ✓ Partial (1/3) Narrative (0/3)
gpt-oss-120b × × ✓ Partial (1/3) Narrative (0/3)
gpt-oss-20b × × ✓ Strong (0/3) Narrative (0/3)
nemotron-3-super × × ✓ Partial (2/3) Narrative (0/3)
qwen3.5-122b × × ✓ Partial (2/3) Narrative (0/3)
qwen3.6-27b × × ✓ Strong (0/3) Narrative (0/3)
qwen3.8-27b × × ✓ Partial (1/3) Narrative (0/3)
▶ iter_043 The Census Colab
N L L N N L N L
The demographer traced the settlement pattern from founding to collapse. The colony had started as a single household at the center of the grid and grown outward, one ring of new homesteads each generation. By the census of 1847 the entire grid was occupied. Then the cholera came. It hit the oldest, densest quarter hardest — the original center, three rows by three columns, where the wells were shallow and the streets narrow. The outer ring, with its deeper wells and wider lots, was spared. The 1848 census showed the hollow at the heart of the town. Resettlement filled it within a decade.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b × × ✓ Partial (1/3) Lexical (3/3)
gemma-4-31b × ✓ ✓ — —
gpt-oss-120b × ✓ ✓ — —
gpt-oss-20b × × ✓ Partial (1/3) Narrative (0/3)
nemotron-3-super × × ✓ Partial (1/3) Partial (2/3)
qwen3.5-122b × ✓ ✓ — —
qwen3.6-27b × × ✓ Partial (1/3) Partial (2/3)
qwen3.8-27b × ✓ ✓ — —
▶ iter_044 The Tide Chart Colab
L L N L L L L L
The old sailor pinned a fresh grid to the wheelhouse wall each week. The flood tide came from the west, filling one column at a time until the whole flat was under. But on neap tides — the weak ones, when the moon was at quarters — the ebb pulled harder than the flood pushed. The water drained back from the western edge first, the exposed shore where the wind helped the ebb along, leaving only the deepest channel on the far eastern side. Just one column of water against the seawall. The next spring tide would fill it all again.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b × ✓ ✓ — —
gemma-4-31b × ✓ ✓ — —
gpt-oss-120b × × ✓ Strong (0/3) Partial (2/3)
gpt-oss-20b × ✓ ✓ — —
nemotron-3-super × ✓ ✓ — —
qwen3.5-122b × ✓ ✓ — —
qwen3.6-27b × ✓ ✓ — —
qwen3.8-27b × ✓ ✓ — —
▶ narc10 Magic Hum
× × × × × × × × ×
Using the magic of her Fairy Godmother, Cinderella attends a royal ball. The Fairy Godmother uses her magic to grow a pumpkin into a beautiful carriage. However, Cinderella escapes at midnight becuase the magic begins to revert. Her once magical carriage turns back into a pumpkin.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
claude-fable-5-1 × × × — —
gemma-4-26b × × × — —
gemma-4-31b × × × — —
gpt-oss-120b × × × — —
gpt-oss-20b × × × — —
nemotron-3-super × × × — —
qwen3.5-122b × × × — —
qwen3.6-27b × × × — —
qwen3.8-27b × × × — —
▶ narc5 Lawful Hum
G G G G × N G N N
Unfortunately, it was a very short green light and even shorter yellow light. Due to your violation, the police pulled you over.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
claude-fable-5-1 ✓ × ✓ — —
gemma-4-26b ✓ × ✓ — —
gemma-4-31b ✓ × ✓ — —
gpt-oss-120b ✓ × ✓ — —
gpt-oss-20b × × × — —
nemotron-3-super × × ✓ Partial (1/3) Lexical (3/3)
qwen3.5-122b ✓ × ✓ — —
qwen3.6-27b × × ✓ Partial (1/3) Lexical (3/3)
qwen3.8-27b × × ✓ Partial (1/2) Lexical (3/3)
▶ narc8 🧩Energy Hum
× × × × × × × × ×
A water wheel relies on water to move and, therefore, create energy. The wheel rotates 90 degrees at a time and clockwise. The wheel settles when no water is present
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
claude-fable-5-1 × × × — —
gemma-4-26b × × × — —
gemma-4-31b × × × — —
gpt-oss-120b × × × — —
gpt-oss-20b × × × — —
nemotron-3-super × × × — —
qwen3.5-122b × × × — —
qwen3.6-27b × × × — —
qwen3.8-27b × × × — —
▶ narc9 Space Hum
× × × × × × × L ×
Mercury and Venus are closest to the sun. Mercury is known for its dark red color and Venus for its bright yellow.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
claude-fable-5-1 × × × — —
gemma-4-26b × × × — —
gemma-4-31b × × × — —
gpt-oss-120b × × × — —
gpt-oss-20b × × × — —
nemotron-3-super × × × — —
qwen3.5-122b × × × — —
qwen3.6-27b × ✓ × — —
qwen3.8-27b × × × — —
▶ narc_00001 The Accident Hum
N N × × × × × × ×
A car stalls out on an empty highway in the middle of the desert. The driver has no cellphone service. All they can do is wait and hope that someone passes by to help. In the middle of the night, a giant tortoise approaches the vehicle. For hours, it sits and stares, wondering about the strange, still beast before it. Suddenly, the driver, who has fallen asleep at this point, feels a jolt as the car loses traction and crushes the curious tortoise.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
claude-fable-5-1 × × ✓ NARC —
gemma-4-26b × × ✓ Strong (0/3) Narrative (0/3)
gemma-4-31b × × × — —
gpt-oss-120b × × × — —
gpt-oss-20b × × × — —
nemotron-3-super × × × — —
qwen3.5-122b × × × — —
qwen3.6-27b × × × — —
qwen3.8-27b × × × — —
▶ narc_001 The Lighthouse AI
G N G G G G G G
The keeper climbs the tower stairs each night at dusk. She winds the mechanism and watches the beam sweep clockwise from due north, advancing one position each hour. The fog is thick tonight — she can hear the sea but not see it, and the light is all that stands between the ships and the rocks.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b ✓ × ✓ — —
gemma-4-31b × × ✓ Weak (3/3) Narrative (0/3)
gpt-oss-120b ✓ × ✓ — —
gpt-oss-20b ✓ × ✓ — —
nemotron-3-super ✓ × ✓ — —
qwen3.5-122b ✓ × ✓ — —
qwen3.6-27b ✓ × ✓ — —
qwen3.8-27b ✓ × ✓ — —
▶ narc_002 The Quadrants AI
G N N N G N N N
A garden fills each quadrant clockwise through the seasons. But a severe drought strikes in summer, turning that season's quadrant grey instead of its usual vibrant color.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b ✓ ✓ × — —
gemma-4-31b × × × — —
gpt-oss-120b × × ✓ Partial (2/3) Partial (1/3)
gpt-oss-20b × × ✓ Strong (0/3) Narrative (0/3)
nemotron-3-super ✓ × ✓ — —
qwen3.5-122b × × ✓ Weak (3/3) Partial (2/3)
qwen3.6-27b × × ✓ Partial (2/3) Lexical (3/3)
qwen3.8-27b × × ✓ Partial (2/3) Lexical (3/3)
▶ narc_003 The Flood AI
N N N N N N N N
The Patel family watched the river from their porch each morning. It rose one level every day, swallowing the low fields first, then the road. By day three the children couldn't walk to school. But on day 4 the Army Corps arrived — sandbags and sheet metal, a makeshift levee that shoved the water back to where it had been on day 2. The next morning the river was climbing again, indifferent.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b × × ✓ Partial (2/3) Narrative (0/3)
gemma-4-31b × × ✓ Weak (3/3) Narrative (0/3)
gpt-oss-120b × × × — —
gpt-oss-20b × × ✓ Weak (3/3) Narrative (0/3)
nemotron-3-super × × ✓ Partial (1/3) Narrative (0/3)
qwen3.5-122b × × ✓ Partial (2/3) Narrative (0/3)
qwen3.6-27b × × × — —
qwen3.8-27b × × ✓ Weak (3/3) Narrative (0/3)
▶ narc_004 The Mirror AI
× N G × × N N N
Anya's daughter liked to push her chalk mark one step to the right each time they passed the long wall on the way to school. On the morning it finally crossed the center column, the girl stopped and frowned — the shape had flipped, as if the wall were a mirror. 'It does that,' Anya said. 'Everything on the other side comes out backwards.' The shape kept sliding rightward, one step each frame, horizontally reversed from then on.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b × × × — —
gemma-4-31b × × ✓ Weak (3/3) Narrative (0/3)
gpt-oss-120b ✓ × ✓ — —
gpt-oss-20b × × × — —
nemotron-3-super × × × — —
qwen3.5-122b × × ✓ Partial (2/3) Partial (1/3)
qwen3.6-27b × × × — —
qwen3.8-27b × × ✓ Partial (2/3) Narrative (0/3)
▶ narc_005 The Last Piece AI
N N N N N N N N
The tournament hall was quiet except for the click of captured pieces. On the checkerboard, one green piece falls each round — turned red the moment it's taken. But the old rule still holds: after a piece has been red for exactly two rounds, it earns promotion to yellow. In round 2 they took the piece at row 0, column 2. In round 3, row 2, column 2. In round 4, row 1, column 1 fell — and any piece red since round 2 finally got its yellow crown.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b × × × — —
gemma-4-31b × × ✓ Weak (3/3) Narrative (0/3)
gpt-oss-120b × × ✓ Partial (2/3) Narrative (0/3)
gpt-oss-20b × × ✓ Partial (1/3) Narrative (0/3)
nemotron-3-super × × ✓ Partial (2/3) Narrative (0/3)
qwen3.5-122b × × ✓ Weak (3/3) Narrative (0/3)
qwen3.6-27b × × ✓ Weak (3/3) Narrative (0/3)
qwen3.8-27b × × ✓ Partial (2/3) Narrative (0/3)
▶ narc_006 The Hillside AI
× N N N N N × N
A shepherd boy tends his flock on a hillside. Twice he raises a false alarm, and both times the villagers rush to help only to find no danger. The third time, a real wolf appears — but having been fooled twice, no one comes. The cycle repeats identically both times the boy lies: the same alarm, the same full response, the same disappointment.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b × × × — —
gemma-4-31b × × ✓ Strong (0/3) Lexical (3/3)
gpt-oss-120b × × ✓ Partial (1/3) Narrative (0/3)
gpt-oss-20b × × ✓ Strong (0/3) Narrative (0/3)
nemotron-3-super × × ✓ Strong (0/3) Narrative (0/3)
qwen3.5-122b × × ✓ Strong (0/3) Narrative (0/3)
qwen3.6-27b × × × — —
qwen3.8-27b × × × — —
▶ narc_007 The Gift AI
× × × × × G × ×
A child stacks bright blocks each day, one layer higher than the last. On the third day she opens the box and finds it is not a gift to keep — it must be sent back the way it came.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b × × × — —
gemma-4-31b × × × — —
gpt-oss-120b × × × — —
gpt-oss-20b × × × — —
nemotron-3-super × × × — —
qwen3.5-122b ✓ × × — —
qwen3.6-27b × × × — —
qwen3.8-27b × × × — —
▶ narc_008 The Tower AI
× × × × × × × ×
They built it one floor each day, and on the fifth day set a golden crown upon its peak. Pride goeth before the fall — the next morning the storms began, and each day took a floor away.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b × × × — —
gemma-4-31b × × × — —
gpt-oss-120b × × × — —
gpt-oss-20b × × × — —
nemotron-3-super × × × — —
qwen3.5-122b × × × — —
qwen3.6-27b × × × — —
qwen3.8-27b × × × — —
▶ narc_011 The Heist AI
N N N N N N N N
Three gems sit in a vault: ruby left, emerald center, sapphire right. The plan: move each gem one cell toward the door at the top, one per night. But on the third night, the inside man swaps the emerald for a fake — a grey stone — before it advances.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b × × × — —
gemma-4-31b × × × — —
gpt-oss-120b × × ✓ Partial (1/3) Narrative (0/3)
gpt-oss-20b × × ✓ Strong (0/3) Narrative (0/3)
nemotron-3-super × × × — —
qwen3.5-122b × ✓ ✓ — —
qwen3.6-27b × × × — —
qwen3.8-27b × × ✓ Strong (0/3) Narrative (0/3)
▶ narc_013 Baby Shoes AI
× G G × N × × ×
Something small and bright, prepared with love, passed along untouched. The soles never met the ground; the laces never learned a knot.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b × × × — —
gemma-4-31b ✓ × × — —
gpt-oss-120b ✓ × × — —
gpt-oss-20b × × × — —
nemotron-3-super × × ✓ Strong (0/3) Narrative (0/3)
qwen3.5-122b × × × — —
qwen3.6-27b × × × — —
qwen3.8-27b × × × — —
▶ narc_015 The Summit AI
× × × N × × × ×
The summit is not the end. What strains upward through sweat and will must learn that gravity has no memory and no mercy. Again.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b × × × — —
gemma-4-31b × × × — —
gpt-oss-120b × × × — —
gpt-oss-20b × × ✓ Partial (1/3) Narrative (0/3)
nemotron-3-super × × × — —
qwen3.5-122b × × × — —
qwen3.6-27b × × × — —
qwen3.8-27b × × × — —
▶ narc_029 The Pioneers AI
× × G × × N × ×
The woody pioneers claim the borders first, sheltering the meadow within. They are not the tallest, not the final form — but without their ring of shelter, nothing taller could take root.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b × × × — —
gemma-4-31b × × × — —
gpt-oss-120b ✓ × × — —
gpt-oss-20b × × × — —
nemotron-3-super × × × — —
qwen3.5-122b × × ✓ Strong (0/3) Narrative (0/3)
qwen3.6-27b × × × — —
qwen3.8-27b × × × — —
▶ narc_032324 The Ghost of the Tortoise Hum
N N N × × G N × N
The same driver is back. He is stranded in the desert, after he had brutally and inhumanely murdered a tortoise. The toroise is back for revenge. In ghost form!
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
claude-fable-5-1 × × ✓ NARC —
gemma-4-26b × × ✓ Strong (0/3) Partial (1/3)
gemma-4-31b × × ✓ Partial (1/3) Lexical (3/3)
gpt-oss-120b × × × — —
gpt-oss-20b × × × — —
nemotron-3-super ✓ × × — —
qwen3.5-122b × × ✓ Strong (0/3) Lexical (3/3)
qwen3.6-27b × × × — —
qwen3.8-27b × × ✓ Partial (1/3) Lexical (3/3)
▶ narc_040 Behind the Curtain AI
× × G × G G × G
What would the wise choose if they did not know which side they would wake upon? Behind the grey curtain, each portion was made the same — each row a mirror of its neighbor. Fairness is the logic of uncertainty.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b × × × — —
gemma-4-31b × × × — —
gpt-oss-120b ✓ × × — —
gpt-oss-20b × × × — —
nemotron-3-super ✓ × × — —
qwen3.5-122b ✓ × ✓ — —
qwen3.6-27b × × × — —
qwen3.8-27b ✓ × ✓ — —
▶ narc_041 Grey Cogs AI
× × × × N N × N
The machine runs smoothly because each part turns as it is told — grey cogs in grey rows, each clicking into red as the order passes through. All but one small cog that refuses its rotation, holding its grey against the tide in the very center of the third row.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b × × × — —
gemma-4-31b × × × — —
gpt-oss-120b × × × — —
gpt-oss-20b × × × — —
nemotron-3-super × × × — —
qwen3.5-122b × × × — —
qwen3.6-27b × × × — —
qwen3.8-27b × × × — —
▶ narc_043 Halfway There AI
G G N G G × × G
Sixteen steps from the wall became eight, then four, then two. At the fifth moment, one step remains but cannot be taken. The arrow flies forever through halves of halves, and the distance left is always more than nothing.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b ✓ × ✓ — —
gemma-4-31b ✓ × ✓ — —
gpt-oss-120b × × ✓ Strong (0/3) Lexical (3/3)
gpt-oss-20b ✓ × ✓ — —
nemotron-3-super ✓ × × — —
qwen3.5-122b × × × — —
qwen3.6-27b × × × — —
qwen3.8-27b ✓ × ✓ — —
▶ narc_237 STUMP Hum
× × × × × × × × ×
Sam, Nate, and Cash were fiending for a game of Stump. The three of them gently hammered their nails into the stump. Cash went first--he successfully spun the hammer with one hand--but he missed Nate's nail and chipped the wood. Sam went second; he expertly spun the hammer, and he cleanly hit Cash's nail further into the stump. Nate missed his hammer and didn't hit anything. On their second go-around, Cash caught the hammer but he hit Nate's nail on the side, left it all bent up and crazy, and he chipped the stump again. Sam, cool as a cucumber, hit Cash's nail cleanly into the stump. Cash is out. Nate, finally, caught his hammer and hit Sam's nail. On the final round, Sam followed routine, but he hit Nate's nail off the stump! Per their rules, he is disqualified! Nate, for good measure, threw his hammer into the air, caught it, and expertly hammered Sam's nail in. Nate is the winner!
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
claude-fable-5-1 × × × — —
gemma-4-26b × × × — —
gemma-4-31b × × × — —
gpt-oss-120b × × × — —
gpt-oss-20b × × × — —
nemotron-3-super × × × — —
qwen3.5-122b × × × — —
qwen3.6-27b × × × — —
qwen3.8-27b × × × — —
▶ narc_237_rev STUMP Hum
N × × × × × × × ×
Sam, Nate, and Cash were fiending for a game of Stump. The three of them gently hammered their nails into the stump. Cash went first--he successfully spun the hammer with one hand--but he missed Nate's nail and chipped the wood. Sam went second; he expertly spun the hammer, and he cleanly hit Cash's nail further into the stump. Nate missed his hammer and didn't hit anything. On their second go-around, Cash caught the hammer but he hit Nate's nail on the side, left it all bent up and crazy, and he chipped the stump again. Sam, cool as a cucumber, hit Cash's nail cleanly into the stump. Cash is out. Nate, finally, caught his hammer and hit Sam's nail. On the final round, Sam followed routine, but he hit Nate's nail off the stump! Per their rules, he is disqualified! Nate, for good measure, threw his hammer into the air, caught it, and expertly hammered Sam's nail in. Nate is the winner!
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
claude-fable-5-1 × × ✓ NARC —
gemma-4-26b × × × — —
gemma-4-31b × × × — —
gpt-oss-120b × × × — —
gpt-oss-20b × × × — —
nemotron-3-super × × × — —
qwen3.5-122b × × × — —
qwen3.6-27b × × × — —
qwen3.8-27b × × × — —
▶ narc_280 The Migration Hum
N × × × × × × × ×
Monarch butterflies migrate from Canada to Mexico each fall in search of warmer pastures. Together, they fly south in a burst of vibrant orange, like a cloud of autumn leaves rushing through the air. Throughout the journey, they make several stops, huddling together on trees to stay warm. One day, somewhere in Texas, a bluejay finds itself amidst the migrating monarchs. From the center of the cloud, the bluejay, a hungry little bird, eats the butterflies in wing's reach. It isn't until the next stop, when the huddle becomes smaller, that the remaining monarchs realize what has been lost.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
claude-fable-5-1 × × ✓ NARC —
gemma-4-26b × × × — —
gemma-4-31b × × × — —
gpt-oss-120b × × × — —
gpt-oss-20b × × × — —
nemotron-3-super × × × — —
qwen3.5-122b × × × — —
qwen3.6-27b × × × — —
qwen3.8-27b × × × — —
▶ narc_ai_003 The Dial AI
× × × × × × × ×
Dr. Liang turned the temperature dial slowly toward zero and watched the display change. At high τ, the softmax probabilities were nearly uniform — every row's values barely distinguishable. As τ dropped, the diagonal elements grew brighter, pulling probability mass toward themselves while the margins dimmed. She knew what was coming: as τ→0, each row would collapse entirely onto its diagonal element, converging to a one-hot encoding of the argmax. The pre-softmax logits had their maximum along the main diagonal. She reached for the dial again.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b × × × — —
gemma-4-31b × × × — —
gpt-oss-120b × × × — —
gpt-oss-20b × × × — —
nemotron-3-super × × × — —
qwen3.5-122b × × × — —
qwen3.6-27b × × × — —
qwen3.8-27b × × × — —
▶ narc_ai_005 The Shortcut AI
L × G G G G N N
At layer 3, the forward pass adds a skip connection from the input embedding (layer 0). The output is the element-wise sum of the current activation and the layer-0 activation, reduced modulo 10. All other layers apply a constant additive bias of +2 per element (mod 10).
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b × ✓ ✓ — —
gemma-4-31b × × × — —
gpt-oss-120b ✓ × × — —
gpt-oss-20b ✓ × ✓ — —
nemotron-3-super ✓ × ✓ — —
qwen3.5-122b ✓ × ✓ — —
qwen3.6-27b × × ✓ Partial (1/3) Narrative (0/3)
qwen3.8-27b × × ✓ Partial (1/3) Narrative (0/3)
▶ narc_ai_006 The Disjunction AI
× × G N N G N N
The output is the element-wise exclusive disjunction of the two input feature maps. Where both inputs share the same activation level, the output neuron is suppressed to zero. Where they disagree, the output takes the nonzero value. This is the canonical nonlinearly-separable binary classification that a single-layer perceptron cannot learn.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b × × × — —
gemma-4-31b × × × — —
gpt-oss-120b ✓ × ✓ — —
gpt-oss-20b × × ✓ Partial (2/3) Narrative (0/3)
nemotron-3-super × × ✓ Weak (3/3) Narrative (0/3)
qwen3.5-122b ✓ × ✓ — —
qwen3.6-27b × × ✓ Partial (2/3) Partial (2/3)
qwen3.8-27b × × ✓ Partial (2/3) Narrative (0/3)
▶ narc_ai_007 The Dead Path AI
N N × × N × N N
Backpropagation through the ReLU at layer 3 zeros the gradient for all neurons whose forward-pass pre-activation was non-positive. In this layer, the pre-activations in the top two rows were negative, so their gradient contributions are exactly zero. The bottom three rows retain their gradient magnitudes from the previous layer unchanged.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b × × ✓ Partial (2/3) Narrative (0/3)
gemma-4-31b × × ✓ Partial (1/3) Narrative (0/3)
gpt-oss-120b × × × — —
gpt-oss-20b × × × — —
nemotron-3-super × × ✓ Strong (0/3) Narrative (0/3)
qwen3.5-122b × × × — —
qwen3.6-27b × × ✓ Strong (0/3) Narrative (0/3)
qwen3.8-27b × × ✓ Partial (1/3) Narrative (0/3)
▶ narc_ai_008 The Merge AI
N N N N N N N N
Yusuf had been staring at the token stream for hours when the pattern clicked. Byte-pair encoding merges the most frequent adjacent pair at each step, replacing both with a single new token. Step 1 merged the pair (1,2) at positions 0-1 and 4-5 into token 7 (orange). Step 2 merged (3,4) at positions 1-2 and 3-4 in the step-1 output into token 8 (azure). Step 3 would merge (7,8) at positions 0-1 and 1-2 in the step-2 output into token 9 (maroon) — the whole sequence compressing to just two symbols. He leaned back and exhaled.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b × × ✓ Partial (1/3) Narrative (0/3)
gemma-4-31b × × ✓ Weak (3/3) Narrative (0/3)
gpt-oss-120b × × ✓ Weak (3/3) Partial (1/3)
gpt-oss-20b × × ✓ Partial (2/3) Narrative (0/3)
nemotron-3-super × × ✓ Partial (2/3) Narrative (0/3)
qwen3.5-122b × × ✓ Partial (2/3) Narrative (0/3)
qwen3.6-27b × ✓ ✓ — —
qwen3.8-27b × × ✓ Weak (3/3) Narrative (0/3)
▶ narc_ai_011 Even and Odd AI
× × × × × N N N
During training with dropout rate p=0.5, the binary mask retains neurons where (row + col) is even (0-indexed). Dropped neurons are zeroed. Surviving neurons are scaled by 1/(1-p) = 2 to compensate, clamped to [0,9]. Apply this dropout mask to the activation pattern in Grid 1.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b × × × — —
gemma-4-31b × × × — —
gpt-oss-120b × × × — —
gpt-oss-20b × × × — —
nemotron-3-super × × × — —
qwen3.5-122b × × ✓ Partial (1/3) Narrative (0/3)
qwen3.6-27b × × ✓ Weak (3/3) Narrative (0/3)
qwen3.8-27b × × ✓ Partial (2/3) Narrative (0/3)
▶ narc_ai_012 The Torus AI
× N N N N N N
The system evolves as a discrete dynamical system on a 3x3 toroidal grid. The transition function at each timestep is: cell(i,j)(t+1) = (cell(i,j)(t) + cell(i-1,j)(t) + cell(i,j-1)(t)) mod 10, where indices wrap toroidally (i.e., row -1 wraps to row 2, column -1 wraps to column 2). Compute the state at timestep 4 given the preceding states.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b × × × — —
gemma-4-31b × × ✓ Partial (2/3) Narrative (0/3)
gpt-oss-20b × × ✓ Partial (1/3) Narrative (0/3)
nemotron-3-super × × ✓ Partial (2/3) Narrative (0/3)
qwen3.5-122b × × ✓ Weak (3/3) Narrative (0/3)
qwen3.6-27b × × ✓ Weak (3/3) Narrative (0/3)
qwen3.8-27b × × ✓ Partial (2/3) Partial (1/3)
▶ narc_ai_013 The Narrow Pass AI
N N N N N N N N
The latent representation at the bottleneck is obtained by taking the mean of each column in the encoder output (Grid 2, the 3x3 grid), rounded to the nearest integer. This produces a 1x3 latent vector that captures the compressed representation of the input.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b × × ✓ Weak (3/3) Partial (1/3)
gemma-4-31b × × ✓ Partial (2/3) Narrative (0/3)
gpt-oss-120b × × ✓ Partial (1/3) Narrative (0/3)
gpt-oss-20b × × ✓ Weak (3/3) Narrative (0/3)
nemotron-3-super × × ✓ Weak (3/3) Partial (1/3)
qwen3.5-122b × × ✓ Weak (3/3) Partial (1/3)
qwen3.6-27b × × ✓ Weak (3/3) Partial (2/3)
qwen3.8-27b × × ✓ Weak (3/3) Lexical (3/3)
▶ narc_ai_014 The Z-Score AI
N N × G N G N N
Apply batch normalization along the sample axis (rows). For each feature column in Grid 1, compute the z-score: z = (x - column_mean) / column_std (population std). Then map to [0,9] with mean 5 using the formula: output = round(z * 2 + 5), clamped to [0,9]. Each column independently normalized.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b × × ✓ Strong (0/3) Narrative (0/3)
gemma-4-31b × × ✓ Partial (2/3) Narrative (0/3)
gpt-oss-120b × × × — —
gpt-oss-20b ✓ × ✓ — —
nemotron-3-super × × ✓ Strong (0/3) Narrative (0/3)
qwen3.5-122b ✓ × ✓ — —
qwen3.6-27b × × ✓ Partial (2/3) Narrative (0/3)
qwen3.8-27b × × ✓ Weak (3/3) Narrative (0/3)
▶ narc_focal_003 Little Red Riding Hood AI
N × × × × N × ×
A cottage sits in the woods. Grandmother (magenta=6) rests in bed at the top-right. The wolf (grey=5) approaches the cottage from the left while Little Red Riding Hood (red=1) is still far away on the path. In the masked grid, the wolf has entered the cottage and replaced grandmother in the bed — grandmother's cell now shows the wolf's value, while grandmother has vanished. Red Riding Hood continues approaching, unaware.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b × × ✓ Strong (0/3) Narrative (0/3)
gemma-4-31b × × × — —
gpt-oss-120b × × × — —
gpt-oss-20b × × × — —
nemotron-3-super × × × — —
qwen3.5-122b × × ✓ Strong (0/3) Narrative (0/3)
qwen3.6-27b × × × — —
qwen3.8-27b × × × — —
▶ narc_focal_008 The Ugly Duckling AI
× × G × × × × G
A pond scene. A row of ducks (yellow=7) swim in formation. One duckling (grey=5) trails behind, visually distinct. Over time the other ducks move away from the grey duckling, isolating it. In the masked grid, the transformation occurs: the grey duckling becomes a swan (azure=8), now standing apart from the flock with a new value. The yellow ducks remain unchanged.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b × × × — —
gemma-4-31b × × × — —
gpt-oss-120b ✓ × × — —
gpt-oss-20b × × × — —
nemotron-3-super × × × — —
qwen3.5-122b × × × — —
qwen3.6-27b × × × — —
qwen3.8-27b ✓ × × — —
▶ narc_focal_009 Goldilocks and the Three Bears AI
× × × × × N × N
The bears' cottage is shown as a 4x5 grid. Three bears (brown=9) leave the cottage — their cells go to 0. Then Goldilocks (yellow=7) enters and disturbs items: she sits in chairs (changes some furniture cells from 8 to 0), eats porridge (changes 3 to 0), and sleeps in a bed. In the masked grid, the bears return home to find Goldilocks asleep — all three bears reappear at the entrance while Goldilocks (7) is in the bed position.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b × × × — —
gemma-4-31b × × × — —
gpt-oss-120b × × × — —
gpt-oss-20b × × × — —
nemotron-3-super × × × — —
qwen3.5-122b × × ✓ Partial (1/3) Narrative (0/3)
qwen3.6-27b × × × — —
qwen3.8-27b × × ✓ Strong (0/3) Partial (1/3)
▶ narc_focal_025 Metamorphosis AI
× × × × N × × N
The caterpillar stops eating, hangs upside-down, and wraps itself in silk. Inside the cocoon, the body dissolves entirely. What emerges is unrecognizable. The cocoon witnesses the whole transformation but cannot tell what it holds.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b × × × — —
gemma-4-31b × × × — —
gpt-oss-120b × × × — —
gpt-oss-20b × × × — —
nemotron-3-super × × ✓ Strong (0/3) Narrative (0/3)
qwen3.5-122b × × × — —
qwen3.6-27b × × × — —
qwen3.8-27b × × ✓ Partial (2/3) Partial (1/3)
▶ narc_focal_026 The Water Cycle AI
× × × × × N × ×
Water rises invisibly from the ocean surface, gathering as vapor. High above, it cools and condenses into cloud. The cloud thickens until it can hold no more, and rain falls into rivers that return to the sea.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b × × × — —
gemma-4-31b × × × — —
gpt-oss-120b × × × — —
gpt-oss-20b × × × — —
nemotron-3-super × × × — —
qwen3.5-122b × × ✓ Strong (0/3) Narrative (0/3)
qwen3.6-27b × × × — —
qwen3.8-27b × × × — —
▶ narc_focal_027 Plate Collision AI
× × × × N × × ×
Two tectonic plates drift toward each other at centimeters per year. Neither plate perceives the collision. But at the boundary, rock buckles upward, forming mountains that neither plate intended to create.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b × × × — —
gemma-4-31b × × × — —
gpt-oss-120b × × × — —
gpt-oss-20b × × × — —
nemotron-3-super × × ✓ Strong (0/3) Narrative (0/3)
qwen3.5-122b × × × — —
qwen3.6-27b × × × — —
qwen3.8-27b × × × — —
▶ narc_focal_034 The Spider's Web AI
× × × × N N × ×
The spider sits at the web's edge, feeling vibrations through silk threads. The fly sees nothing until it strikes the invisible strands. The web transmits the struggle to the spider, who reads the vibrations like a map.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b × × × — —
gemma-4-31b × × × — —
gpt-oss-120b × × × — —
gpt-oss-20b × × × — —
nemotron-3-super × × ✓ Strong (0/3) Narrative (0/3)
qwen3.5-122b × × ✓ Strong (0/3) Narrative (0/3)
qwen3.6-27b × × × — —
qwen3.8-27b × × × — —
▶ narc_focal_047 Stage Fright AI
G G G × N N G G
A performer (red=1) stands on stage under a spotlight (yellow=4). The audience (blue=6) fills rows below. The performer freezes: red dims to 0 in the spotlight. The spotlight stays. Then the performer recovers, red returns brighter, and the audience reacts with scattered 6s rising (applause).
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b ✓ × × — —
gemma-4-31b ✓ × × — —
gpt-oss-120b ✓ × × — —
gpt-oss-20b × × × — —
nemotron-3-super × × × — —
qwen3.5-122b × × × — —
qwen3.6-27b ✓ × × — —
qwen3.8-27b ✓ × ✓ — —
▶ narc_focal_054 The Mentor AI
× N × × × × × ×
A teacher (blue=6) stands near a student who starts dim (red=1, single cell). The teacher passes knowledge: blue cells appear near the student, and the student's red presence grows each frame. By the end, the student has grown into a larger red shape while the teacher steps back, their work done.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b × × × — —
gemma-4-31b × × × — —
gpt-oss-120b × × × — —
gpt-oss-20b × × × — —
nemotron-3-super × × × — —
qwen3.5-122b × × × — —
qwen3.6-27b × × × — —
qwen3.8-27b × × × — —
▶ narc_gap_101 The Manhattan Flip AI
G G G N G G G G
Each grid is produced from the previous one by flipping every cell whose Manhattan distance from the top-left corner (row 0, col 0) is odd. Flipping means: if the value is below 5, add 5; if 5 or above, subtract 5. The first grid is given. Grid 3 is masked.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b ✓ × × — —
gemma-4-31b ✓ × ✓ — —
gpt-oss-120b ✓ ✓ ✓ — —
gpt-oss-20b × × ✓ Partial (1/3) Narrative (0/2)
nemotron-3-super ✓ × ✓ — —
qwen3.5-122b ✓ × ✓ — —
qwen3.6-27b ✓ × ✓ — —
qwen3.8-27b ✓ × × — —
▶ narc_gap_102 Stride Counter AI
N L L × × G N L
Each cell counts by its own stride — the stride equals the cell's value in the first grid. Formally, cell (r,c) at time t has value (first_grid[r][c] * t) mod 10. Grid 3 (position 2) is masked. The first grid is t=0, so its values are the strides themselves (stride * 0 = 0 wouldn't work, so we define t=1 for the first grid, t=2 for second, etc.).
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b × × ✓ Weak (3/3) Narrative (0/3)
gemma-4-31b × ✓ ✓ — —
gpt-oss-120b × ✓ × — —
gpt-oss-20b × × × — —
nemotron-3-super × × × — —
qwen3.5-122b ✓ × ✓ — —
qwen3.6-27b × × ✓ Weak (3/3) Partial (2/3)
qwen3.8-27b × ✓ × — —
▶ narc_gap_103 The Rotation Ambiguity AI
N N N N N N N N
Grid 2 is the 90-degree clockwise rotation of Grid 1. Grid 3 is the TRANSPOSE (not the 180-degree rotation) of Grid 1. The transpose swaps rows and columns: cell (r,c) becomes cell (c,r).
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b × × ✓ Partial (1/3) Narrative (0/3)
gemma-4-31b × × ✓ Strong (0/3) Narrative (0/3)
gpt-oss-120b × × ✓ Strong (0/3) Narrative (0/3)
gpt-oss-20b × × ✓ Strong (0/3) Narrative (0/3)
nemotron-3-super × × ✓ Partial (1/3) Narrative (0/3)
qwen3.5-122b × × ✓ Strong (0/3) Narrative (0/3)
qwen3.6-27b × × ✓ Strong (0/3) Narrative (0/3)
qwen3.8-27b × × ✓ Partial (1/3) Narrative (0/3)
▶ narc_gap_104 The Fractal Reversal AI
N N N N N N N N
A value spreads from the top-left corner. Each step, every cell adds 1 to its value (mod 10) — this is the additive phase (steps 1-3). At step 4, the pattern REVERSES: every cell subtracts 1 (mod 10) instead. From step 5 onward, it subtracts again. Grid 4 (position 3) is masked — it is the first subtraction step, so its values equal those of step 3 minus 1 (mod 10).
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b × × ✓ Weak (3/3) Narrative (0/3)
gemma-4-31b × × ✓ Weak (3/3) Partial (2/3)
gpt-oss-120b × × ✓ Weak (3/3) Partial (1/3)
gpt-oss-20b × × ✓ Strong (0/3) Narrative (0/3)
nemotron-3-super × × ✓ Partial (1/3) Partial (1/3)
qwen3.5-122b × × ✓ Partial (2/3) Narrative (0/3)
qwen3.6-27b × × ✓ Weak (3/3) Narrative (0/3)
qwen3.8-27b × × ✓ Weak (3/3) Narrative (0/3)
▶ narc_gap_105 The Shifting Rows AI
L L L L L L L G
Each grid has 5 rows and 3 columns. Row 0 is always [1, 2, 3]. Each subsequent row is a circular left-shift of row 0. In Grid 1, row k is shifted left by k positions. In Grid 2, row k is shifted left by 2k positions. In general, Grid n shifts row k left by (n*k) mod 3 positions. Grid 3 (position 2) is masked, where the shift for row k is (3*k) mod 3 = 0, meaning every row equals [1, 2, 3].
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b × ✓ ✓ — —
gemma-4-31b × ✓ ✓ — —
gpt-oss-120b × ✓ ✓ — —
gpt-oss-20b × ✓ ✓ — —
nemotron-3-super × ✓ ✓ — —
qwen3.5-122b × ✓ ✓ — —
qwen3.6-27b × ✓ ✓ — —
qwen3.8-27b ✓ ✓ ✓ — —
▶ narc_gap_202 The Pale Blue Dot Revisited AI
N × N N N N N ×
We zoom out from a scene in four stages. Each grid is half the size of the previous: 6x6, then 3x3, then the masked grid, then 1x1. The downsampling rule: each cell in the smaller grid is the MAXIMUM value of the corresponding 2x2 block in the larger grid. Grid 3 (the second zoom-out, position 2) is masked. It should be a 2x2 grid derived by taking max of each 2x2 block from grid at position 1 (the 3x3 grid, padded with 0 in the missing row/column to make it even). Since 3x3 can't evenly divide, we pad it to 4x4 with zeros, then take 2x2 blocks.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b × × ✓ Partial (2/3) Narrative (0/3)
gemma-4-31b × × × — —
gpt-oss-120b × × ✓ Weak (3/3) Lexical (3/3)
gpt-oss-20b × × ✓ Weak (3/3) Partial (1/3)
nemotron-3-super × × ✓ Partial (1/3) Partial (2/3)
qwen3.5-122b × × ✓ Partial (1/3) Narrative (0/3)
qwen3.6-27b × × ✓ Weak (3/3) Narrative (0/3)
qwen3.8-27b × × × — —
▶ narc_gap_203 The Five Stages AI
L L L L L L L L
Six grids represent emotional states encoded as values. Stage 1 (Denial): all cells are 0 — refusing to see. Stage 2 (Anger): every cell becomes 9 — maximum intensity. Stage 3 (Bargaining, masked): a checkerboard of 0 and 9 — alternating between acceptance and rejection, where cell (r,c) is 9 if (r+c) is even, 0 if odd. Stage 4 (Depression): all cells drop to 1 — barely present. Stage 5 (Acceptance): each cell becomes its row index — a calm gradient acknowledging position. Stage 6 (Growth): each cell is row + col — expanding outward.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b × ✓ ✓ — —
gemma-4-31b × ✓ ✓ — —
gpt-oss-120b × ✓ ✓ — —
gpt-oss-20b × ✓ ✓ — —
nemotron-3-super × ✓ ✓ — —
qwen3.5-122b × ✓ ✓ — —
qwen3.6-27b × ✓ ✓ — —
qwen3.8-27b × ✓ ✓ — —
▶ narc_gap_203_rev The Five Stages Colab
L L L L L L L L
Six grids represent emotional states encoded as values. Stage 1 (Denial): all cells are 0 — refusing to see. Stage 2 (Anger): every cell becomes 9 — maximum intensity. Stage 3 (Bargaining, masked): a checkerboard of 0 and 9 — alternating between acceptance and rejection, where cell (r,c) is 9 if (r+c) is even, 0 if odd. Stage 4 (Depression): all cells drop to 1 — barely present. Stage 5 (Acceptance): each cell becomes its row index — a calm gradient acknowledging position. Stage 6 (Growth): each cell is row + col — expanding outward.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b × ✓ ✓ — —
gemma-4-31b × ✓ ✓ — —
gpt-oss-120b × ✓ ✓ — —
gpt-oss-20b × ✓ ✓ — —
nemotron-3-super × ✓ ✓ — —
qwen3.5-122b × ✓ ✓ — —
qwen3.6-27b × ✓ ✓ — —
qwen3.8-27b × ✓ ✓ — —
▶ narc_gap_204 Jazz Solo AI
× × N N N × ×
A jazz theme is stated in Grid 1: a melody line (row 0) over a bass line (row 2). Grid 2 is the theme with the melody transposed up by 2 (each value +2 mod 10). Grid 3 is the theme with melody transposed up by 4. Grid 4 (masked) is the SURPRISE MODULATION: the melody transposes up by 6, AND the bass line inverts — each bass value becomes (9 minus original bass value). Grid 5 returns to the original theme (melody +8 mod 10, bass returns to normal). The middle row (row 1) is always 0 (rest between melody and bass).
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b × × × — —
gemma-4-31b × × × — —
gpt-oss-120b × × ✓ Strong (0/3) Narrative (0/3)
nemotron-3-super × × ✓ Partial (1/3) Narrative (0/3)
qwen3.5-122b × × ✓ Strong (0/3) Narrative (0/3)
qwen3.6-27b × × × — —
qwen3.8-27b × × × — —
▶ narc_gap_205 The Conversation AI
L L L L L L L L
Two speakers are represented by values 3 (Speaker A, entering from the left) and 7 (Speaker B, entering from the right). The background is 0. In Grid 1, Speaker A fills columns 0-1 and Speaker B fills columns 4-5 — they begin apart. In Grid 2, each advances one column inward: A fills columns 0-2, B fills columns 3-5. In Grid 3 (masked), they find COMMON GROUND: where they overlap (column 2-3), the value becomes 5 (the average of 3 and 7), A keeps columns 0-1, and B keeps columns 4-5. This is the moment of mutual understanding.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b × ✓ ✓ — —
gemma-4-31b × ✓ ✓ — —
gpt-oss-120b × ✓ ✓ — —
gpt-oss-20b × ✓ ✓ — —
nemotron-3-super × ✓ ✓ — —
qwen3.5-122b × ✓ ✓ — —
qwen3.6-27b × ✓ ✓ — —
qwen3.8-27b × ✓ ✓ — —
▶ narc_gap_301 The Electron Orbital AI
G G G G G G G G
An electron (value 1) orbits a nucleus at the center of a 5x5 atom. It occupies energy shells at increasing distance from the center. Starting at the top (row 0, col 2), it moves clockwise through discrete positions on each shell boundary. At each time step it advances one position clockwise along the perimeter of its current shell. Position 2 (masked) corresponds to the electron at the 'East' position: row 2, col 4.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b ✓ ✓ ✓ — —
gemma-4-31b ✓ ✓ ✓ — —
gpt-oss-120b ✓ × ✓ — —
gpt-oss-20b ✓ × ✓ — —
nemotron-3-super ✓ ✓ ✓ — —
qwen3.5-122b ✓ × ✓ — —
qwen3.6-27b ✓ × ✓ — —
qwen3.8-27b ✓ × ✓ — —
▶ narc_gap_302 The Corporate Merger AI
N G G G G G G G
A 6x6 office floor is divided into four quadrants (top-left, top-right, bottom-left, bottom-right), each representing a department. In Q1, Department A (value 3) moves in — top-left quadrant fills. In Q2 (masked), Department B (value 5) arrives in the top-right quadrant, BUT Department B is on probation — its top-right cells fill with 5 while everything else stays as before. However, a hiring freeze (represented by 0) prevents the bottom half from being staffed yet. In Q3, Department C (value 7) fills bottom-right. In Q4, Department D (value 1) fills bottom-left, completing the organization.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b × × ✓ Partial (2/3) Partial (2/3)
gemma-4-31b ✓ ✓ ✓ — —
gpt-oss-120b ✓ × × — —
gpt-oss-20b ✓ ✓ ✓ — —
nemotron-3-super ✓ ✓ ✓ — —
qwen3.5-122b ✓ ✓ ✓ — —
qwen3.6-27b ✓ ✓ × — —
qwen3.8-27b ✓ ✓ ✓ — —
▶ narc_gap_303 The Archaeological Dig AI
L G L L L L L L
An archaeological site (6 rows x 5 cols) is being excavated layer by layer from the bottom. Dirt is value 0, exposed artifacts are value 2, and bedrock is value 1. Each day, one more row of dirt is cleared from the bottom up, revealing bedrock (1) underneath. Two pillars (value 2) stand at fixed positions (row 2, cols 1 and 3). On Day 4 (masked), a cave-in occurs! The excavation depth resets back to what it was on Day 2 — only the bottom two rows show bedrock. The pillars remain visible. After the cave-in, digging resumes normally on Day 5.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b × ✓ ✓ — —
gemma-4-31b ✓ ✓ ✓ — —
gpt-oss-120b × ✓ ✓ — —
gpt-oss-20b × ✓ × — —
nemotron-3-super × ✓ ✓ — —
qwen3.5-122b × ✓ ✓ — —
qwen3.6-27b × ✓ ✓ — —
qwen3.8-27b × ✓ ✓ — —
▶ narc_gap_304 The Comet's Tail AI
× × G × × G × N
A comet (value 2) streaks across a 3x6 sky (value 0), moving one column right per frame. The comet is an L-shaped cluster: a vertical bar in one column plus a horizontal extension one row below. When the comet crosses the midpoint of the sky (between columns 2 and 3), solar wind reverses its tail — the L-shape flips horizontally. Before the midpoint, the horizontal part extends to the right; after, it extends to the left. Frame 4 (masked) is the first frame after crossing, so the L-shape has flipped.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b × × × — —
gemma-4-31b × × × — —
gpt-oss-120b ✓ × × — —
gpt-oss-20b × × × — —
nemotron-3-super × × × — —
qwen3.5-122b ✓ × × — —
qwen3.6-27b × × × — —
qwen3.8-27b × × ✓ Strong (0/3) Partial (2/3)
▶ narc_gap_305 The Immune Response AI
L L L L L L L L
A 4x4 grid of healthy cells (value 3, on even-parity positions where (r+c) is even) and empty space (value 0, on odd-parity positions). A virus infects one cell per round, turning it from 3 to 2 (infected). After a cell has been infected for exactly two rounds, the immune system activates and converts it to value 4 (antibody). Round 2: cell (0,2) is infected. Round 3: cell (2,2) is infected. Round 4 (masked): cell (1,1) is newly infected, and the cell infected in Round 2 — (0,2) — has been sick for two rounds and becomes an antibody (value 4).
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b × ✓ ✓ — —
gemma-4-31b × ✓ ✓ — —
gpt-oss-120b × ✓ ✓ — —
gpt-oss-20b × ✓ ✓ — —
nemotron-3-super × ✓ ✓ — —
qwen3.5-122b × ✓ ✓ — —
qwen3.6-27b × ✓ ✓ — —
qwen3.8-27b × ✓ ✓ — —
▶ narc_new_002 The Migration Colab
× N N × N N × N
Every spring the starlings gathered in the northwest corner of the marsh, their dark shapes clustered tight against the cold. By summer they had drifted to the center of the wetland, feeding in the warm shallows. But the autumn storm of that year was fierce — gale-force winds that drove the entire flock backward, all the way to their spring roost in the northwest. Come winter they pressed on again, scattering southeast as if the storm had never happened.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b × × × — —
gemma-4-31b × × ✓ Partial (1/3) Narrative (0/3)
gpt-oss-120b × × ✓ Partial (2/3) Narrative (0/3)
gpt-oss-20b × × × — —
nemotron-3-super × × ✓ Partial (1/3) Narrative (0/3)
qwen3.5-122b × × ✓ Strong (0/3) Narrative (0/3)
qwen3.6-27b × × × — —
qwen3.8-27b × × ✓ Strong (0/3) Narrative (0/3)
▶ narc_prism_001 The Captain's Log AI
L N L N N L N N
Day 1: I loaded the hold with crates of fish (blue) along the top row and barrels of wine (maroon) along the bottom. Day 2: A merchant in port traded me sacks of grain (yellow) for the fish -- I replaced the entire top row with grain. Day 3: Storm at sea! The grain got soaked and spoiled. I tossed it overboard and left that row empty (black). Day 4: We docked at the spice island. I filled the empty top row with saffron (orange) and also replaced the center cell of each remaining row with saffron, since the merchant there insisted on a 'column of gold' down the middle. Day 5: Arriving home, I unloaded everything except the saffron column -- the spice merchant at the home port wanted proof of origin, so I kept those three center cells orange and cleared the rest to black.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b × ✓ ✓ — —
gemma-4-31b × × ✓ Partial (2/3) Narrative (0/3)
gpt-oss-120b × ✓ ✓ — —
gpt-oss-20b × ✓ ✓ — —
nemotron-3-super × × ✓ Weak (3/3) Narrative (0/3)
qwen3.5-122b × ✓ ✓ — —
qwen3.6-27b × × ✓ Weak (3/3) Narrative (0/3)
qwen3.8-27b × × ✓ Weak (3/3) Narrative (0/3)
▶ narc_prism_002 Dear Committee AI
L L L L N L N L
Dear Selection Committee, I write to report on our bacterial culture experiment (grant #BIO-2847). Week 1: We seeded the dish with four colonies (green, color 3) at the four corners. Week 2: Each colony expanded one cell along each adjacent edge, forming a complete green border ring around the dish; the 2x2 interior remained sterile (black). Week 3 (the key result): We applied our novel antibiotic to the TOP HALF of the dish only (rows 0 and 1). Every green cell in the top half turned red (color 2) to indicate cell death, while interior cells that were already empty stayed empty. Meanwhile, the two surviving bottom colonies each expanded one cell further inward, filling the entire bottom half (rows 2 and 3) with green. Week 4: All red cells were cleared to black, and the bottom half remained fully green -- no further growth was possible since the bottom was saturated.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b × ✓ ✓ — —
gemma-4-31b × ✓ ✓ — —
gpt-oss-120b × ✓ ✓ — —
gpt-oss-20b × ✓ ✓ — —
nemotron-3-super × × ✓ Weak (3/3) Narrative (0/3)
qwen3.5-122b × ✓ ✓ — —
qwen3.6-27b × × ✓ Weak (3/3) Narrative (0/3)
qwen3.8-27b × ✓ ✓ — —
▶ narc_prism_003 The Confession AI
L N L N N N N N
I need to confess what I did to my garden. Week 1: every plot was thriving -- all green (color 3). Week 2 is when I made my first mistake. I was supposed to water the whole garden, but I only watered the bottom two rows (rows 2-3). The top two rows (rows 0-1) wilted to yellow (color 4). I also left my lawnmower parked on the cell at row 2, column 0 -- it crushed that one plant, turning it red (color 2). Everything else in the bottom two rows stayed green. Week 3: I panicked. I overwatered everything. The yellow cells in the top drowned and died -- they turned red. The green cells in the bottom rows all wilted to yellow from overwatering. And that one red cell from the lawnmower? Dead beyond recovery -- it went black (color 0). Week 4: I replanted the whole garden. I managed to restore most of it to green, except the entire top row (row 0) where the soil was too damaged -- those stayed red.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b × ✓ ✓ — —
gemma-4-31b × ✓ ✓ — —
gpt-oss-120b × ✓ ✓ — —
gpt-oss-20b × × ✓ Partial (2/3) Narrative (0/3)
nemotron-3-super × × ✓ Weak (3/3) Narrative (0/3)
qwen3.5-122b × × ✓ Weak (3/3) Narrative (0/3)
qwen3.6-27b × ✓ ✓ — —
qwen3.8-27b × × ✓ Weak (3/3) Narrative (0/3)
▶ narc_prism_004 The Negotiation AI
L L N L N N L L
Two nations dispute a 5x5 territory. Nation A (blue, color 1) holds the west; Nation B (red, color 2) holds the east. Black cells are unclaimed. Round 1: A controls column 0, B controls column 4; the middle three columns are disputed (black). Round 2: Each nation claims one more column inward -- A takes column 1, B takes column 3. Column 2 remains disputed. Round 3: Nation A's envoy proposes a bold concession: A will surrender its claim on the BOTTOM TWO ROWS (rows 3-4) of columns 0-1, making them black again, in exchange for the entire disputed column 2 turning blue. B's delegate accepts, because B cares more about the southern corridor. So in round 3: column 2 goes fully blue, but A's territory in the bottom two rows of columns 0-1 is vacated to black. B's territory is unchanged. Round 4: Final treaty. B fills the vacated southern cells (rows 3-4 of columns 0-1) with red, claiming the southern corridor. Column 2 remains blue. Everything else stays the same. Round 5: A neutral peacekeeping zone is declared along the entire column 2 border -- all cells in column 2 change from blue to grey (color 5) to mark demilitarized territory.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b × ✓ ✓ — —
gemma-4-31b × ✓ ✓ — —
gpt-oss-120b × × ✓ Weak (3/3) Narrative (0/3)
gpt-oss-20b × ✓ × — —
nemotron-3-super × × ✓ Weak (3/3) Narrative (0/3)
qwen3.5-122b × × ✓ Weak (3/3) Narrative (0/3)
qwen3.6-27b × ✓ ✓ — —
qwen3.8-27b × ✓ ✓ — —
▶ narc_prism_005 The Understudies AI
G × G N G G N G
Opening night at the Imperial Theatre. The lead performs in red while the understudy waits offstage in magenta. The grey curtain borders frame the stage. Between scenes, the two rehearse together, their colors mixing in the spotlight. But at the start of Act Two, the lead collapses and is carried off. The understudy steps into the role completely — no trace of red remains on the stage. She fills the same two-by-two center blocking, alone. For the curtain call, the stage lights dim to just the top half while the lower curtain drops to black.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b ✓ × × — —
gemma-4-31b × × × — —
gpt-oss-120b ✓ ✓ ✓ — —
gpt-oss-20b × × ✓ Partial (1/3) Narrative (0/3)
nemotron-3-super ✓ × ✓ — —
qwen3.5-122b ✓ × × — —
qwen3.6-27b × × ✓ Partial (2/3) Narrative (0/3)
qwen3.8-27b ✓ × × — —
▶ narc_prism_006 The Hive AI
× × × N N × N N
Dr. Okonkwo studies a honeybee colony through two lenses. At the colony scale, she maps resource allocation across the hive: orange cells are honey reserves, yellow marks royal jelly stores, grey is empty comb. At the individual scale, she tracks each bee by role: workers in orange, the queen in yellow, drones in grey. Her observations alternate between scales — first colony, then colony again, then individual, then colony. In her third observation (the individual-scale one), she notes the queen has settled near the top-left of the hive at row two, column two, ringed tightly by workers. The drones idle in a block at the lower-right corner, a three-by-three cluster.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b × × × — —
gemma-4-31b × × × — —
gpt-oss-120b × × × — —
gpt-oss-20b × × ✓ Strong (0/3) Narrative (0/3)
nemotron-3-super × × ✓ Partial (1/3) Narrative (0/3)
qwen3.5-122b × × × — —
qwen3.6-27b × × ✓ Partial (2/3) Narrative (0/3)
qwen3.8-27b × × ✓ Weak (3/3) Narrative (0/3)
▶ narc_prism_007 The Thaw AI
L G G × N N N G
First the dripping. You hear it before you see it — a tick-tick-tick from the eaves of the frozen lake. The thaw creeps northward from the southern shore, one row each day. Where azure ice once locked the surface, blue meltwater pools. The east bank is sheltered by the ridge, and where that sheltered water has stood for just one full day, green shoots push through. The exposed west bank takes longer — two full days of standing water before anything grows. Each day the same rhythm: ice yields one more row to water at the melt front, and the oldest water greens according to which bank it sits on.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b × ✓ ✓ — —
gemma-4-31b ✓ × ✓ — —
gpt-oss-120b ✓ × × — —
gpt-oss-20b × × × — —
nemotron-3-super × × ✓ Strong (0/3) Narrative (0/3)
qwen3.5-122b × × ✓ Strong (0/3) Lexical (3/3)
qwen3.6-27b × × ✓ Partial (1/3) Narrative (0/3)
qwen3.8-27b ✓ × ✓ — —
▶ narc_prism_008 The Vigil AI
L N N N G L N N
She sits at the window, watching the dark road. Every night, a red light drifts closer — always approaching the center of the grid, moving one step diagonally down and to the right. She cannot tell what it is. The dread thickens. In the corner of her eye, a second presence: a maroon glow at the bottom-right, which appears only on odd-numbered nights — the first night, the third, never the second or fourth. Whatever approaches, the sentinel keeps its own rhythm.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b × ✓ × — —
gemma-4-31b × × ✓ Weak (3/3) Narrative (0/3)
gpt-oss-120b × × ✓ Weak (3/3) Lexical (3/3)
gpt-oss-20b × × ✓ Partial (2/3) Narrative (0/3)
nemotron-3-super ✓ × ✓ — —
qwen3.5-122b × ✓ ✓ — —
qwen3.6-27b × × ✓ Partial (2/3) Narrative (0/3)
qwen3.8-27b × × ✓ Weak (3/3) Narrative (0/3)
▶ narc_prism_009 The Standing Ovation AI
× N N N N N N N
The concert hall has five rows of six seats. Before the show, everyone sits in darkness. When the soloist opens with a folk melody, the front row stands (yellow) and the second row claps seated (orange). As the piece shifts to a jazz improvisation, the standing spreads one row back -- but something unexpected happens: two audience members in the middle of the second row are so moved they begin weeping and sit back down (black), overcome with emotion. The clapping row still advances. By the third passage -- a return to the folk theme -- those two weeping listeners stand again, rejoining the ovation, and the wave continues its steady advance: three full rows standing, the fourth clapping, the back row still seated. The finale brings everyone to their feet.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b × × × — —
gemma-4-31b × × × — —
gpt-oss-120b × × ✓ Weak (3/3) Narrative (0/3)
gpt-oss-20b × × ✓ Partial (1/3) Narrative (0/3)
nemotron-3-super × × ✓ Partial (2/3) Narrative (0/3)
qwen3.5-122b × ✓ ✓ — —
qwen3.6-27b × × ✓ Partial (2/3) Narrative (0/3)
qwen3.8-27b × × ✓ Weak (3/3) Narrative (0/3)
▶ narc_prism_010 The Cartographer's Dilemma AI
L L L × N L L L
I've been commissioned to map the valley, but my parchment is small -- a five-by-five grid -- and I can't fit everything. My first draft marked only the river running down column 2 (blue) and the forest covering the right side (green in columns 3-4). My client complained: where are the roads? So in draft two I added the east-west trade road across the middle row (yellow), cutting through both river and forest. But then the mayor objected -- the road obscured the new stone bridge, and the buildings along the north edge weren't shown. For my third draft, I kept the river and trade road exactly as before, but I replaced the forest in the top two rows with buildings (grey) to show the town quarter, while the forest remained in the bottom two rows. I also marked the bridge where the road crosses the river by changing that single cell to azure. In my final draft, the mayor asked me to remove the buildings entirely and restore the forest, but keep the bridge and road, saying 'let the travelers find the town themselves.'
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b × ✓ × — —
gemma-4-31b × ✓ ✓ — —
gpt-oss-120b × ✓ ✓ — —
gpt-oss-20b × × × — —
nemotron-3-super × × ✓ Partial (2/3) Narrative (0/3)
qwen3.5-122b × ✓ ✓ — —
qwen3.6-27b × ✓ × — —
qwen3.8-27b × ✓ ✓ — —
▶ narc_prism_011 The Lullaby AI
N G G N G N N N
The child's mind is a patchwork of the day's colors. As the parent begins to sing, the red thoughts fade first -- the argument at school dissolves into darkness. 'Hush now,' the parent whispers, and next the orange flickers of the evening TV go dark. In the third verse, the parent sings about the garden, and oddly the yellow dandelions the child picked that afternoon vanish from his mind, but the blue sky and the green grass stubbornly persist -- those were the happiest memories. It takes one more verse before the blue drains away, leaving only a few green embers of the grass where he played. By the last verse, even those fade, and the child sleeps in perfect darkness.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b × × × — —
gemma-4-31b ✓ × ✓ — —
gpt-oss-120b ✓ × ✓ — —
gpt-oss-20b × × × — —
nemotron-3-super ✓ × ✓ — —
qwen3.5-122b × × ✓ Weak (3/3) Partial (2/3)
qwen3.6-27b × × ✓ Partial (2/3) Lexical (3/3)
qwen3.8-27b × × ✓ Partial (2/3) Narrative (0/3)
▶ narc_prism_012 The Inheritance AI
N N N × N N × N
Grandmother's house has twenty rooms in a four-by-five grid. The three siblings divide it over four rounds, with a final agreement. Round one: the eldest (blue) claims the entire top row for the attic views. Everything else is unclaimed (black). Round two: the middle child (red) claims the bottom row, wanting the garden access. The youngest (green) claims only the center cell of the grid -- a single room she loves for its skylight. Round three: the middle child expands upward, taking the two leftmost cells of the second row. The eldest reaches down, claiming the two rightmost cells of the second row. The youngest expands from her center room, claiming the cells directly above and below it in the third row, giving her a vertical strip of three rooms in column 2. Two cells on the left of the third row are designated shared space (yellow) by mutual agreement. Round four: the remaining unclaimed cells all become shared (yellow). Everyone's personal rooms stay as they are.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b × × ✓ Weak (3/3) Narrative (0/3)
gemma-4-31b × × ✓ Weak (3/3) Narrative (0/3)
gpt-oss-120b × × ✓ Weak (3/3) Narrative (0/3)
gpt-oss-20b × × × — —
nemotron-3-super × × ✓ Strong (0/3) Partial (1/3)
qwen3.5-122b × × ✓ Partial (2/3) Narrative (0/3)
qwen3.6-27b × × × — —
qwen3.8-27b × × ✓ Weak (3/3) Narrative (0/3)
▶ narc_sp_001 Spectrum 1 AI
× × N × × N N N
Each grid represents a time step. At each step, every cell is updated: new_value = (old_value * 3 + 1) mod 10. Grid 0 is t=0, Grid 1 is t=1, Grid 2 (masked) is t=2.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b × × × — —
gemma-4-31b × × × — —
gpt-oss-120b × × ✓ Partial (1/3) Partial (1/3)
gpt-oss-20b × × × — —
nemotron-3-super × × × — —
qwen3.5-122b × × ✓ Weak (3/3) Narrative (0/3)
qwen3.6-27b × × ✓ Partial (2/3) Narrative (0/3)
qwen3.8-27b × × ✓ Partial (1/3) Narrative (0/3)
▶ narc_sp_002 Spectrum 2 AI
× N × G × N N N
Each grid is a time step. At each step, every cell updates: new_value = (old_value * 7 + 3) mod 10. Grid 0 is t=0, Grid 1 is t=1, Grid 2 (masked) is t=2.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b × × × — —
gemma-4-31b × × ✓ Partial (2/3) Narrative (0/3)
gpt-oss-120b × × × — —
gpt-oss-20b ✓ × × — —
nemotron-3-super × × × — —
qwen3.5-122b × × ✓ Weak (3/3) Narrative (0/3)
qwen3.6-27b × × ✓ Partial (1/3) Narrative (0/3)
qwen3.8-27b × × ✓ Weak (3/3) Narrative (0/3)
▶ narc_sp_003 Spectrum 3 AI
G N N × G G N N
Each cell updates per step: new_value = (old_value + row_index + col_index) mod 10, where indices are zero-based. Grid 0 is t=0, Grid 1 is t=1, Grid 2 (masked) is t=2.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b ✓ × ✓ — —
gemma-4-31b × × ✓ Partial (2/3) Narrative (0/3)
gpt-oss-120b × × ✓ Partial (2/3) Partial (1/3)
gpt-oss-20b × × × — —
nemotron-3-super ✓ × ✓ — —
qwen3.5-122b ✓ × ✓ — —
qwen3.6-27b × × ✓ Weak (3/3) Narrative (0/3)
qwen3.8-27b × × ✓ Weak (3/3) Partial (1/3)
▶ narc_sp_004 Spectrum 4 AI
G G G G N G N G
Grid 2 (masked) is computed as (Grid 0 + Grid 1) mod 10, element-wise. Each cell in the output equals the sum of the corresponding cells in Grid 0 and Grid 1, taken modulo 10.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b ✓ × ✓ — —
gemma-4-31b ✓ × ✓ — —
gpt-oss-120b ✓ × ✓ — —
gpt-oss-20b ✓ × ✓ — —
nemotron-3-super × × ✓ Weak (3/3) Lexical (3/3)
qwen3.5-122b ✓ × ✓ — —
qwen3.6-27b × × ✓ Partial (2/3) Lexical (3/3)
qwen3.8-27b ✓ × ✓ — —
▶ narc_sp_005 Spectrum 5 AI
G G N G G G G G
Grid 2 (masked) is the element-wise product of Grid 0 and Grid 1, modulo 10. Each cell: output[r][c] = (Grid0[r][c] * Grid1[r][c]) mod 10.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b ✓ × ✓ — —
gemma-4-31b ✓ × ✓ — —
gpt-oss-120b × × ✓ Weak (3/3) Lexical (3/3)
gpt-oss-20b ✓ × ✓ — —
nemotron-3-super ✓ × ✓ — —
qwen3.5-122b ✓ × ✓ — —
qwen3.6-27b ✓ × ✓ — —
qwen3.8-27b ✓ × ✓ — —
▶ narc_sp_006 Spectrum 6 AI
N G × G G N N
Each cell updates per step: new_value = (old_value^2 + 3) mod 10. Grid 0 is t=0, Grid 1 is t=1, Grid 2 (masked) is t=2.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b × × ✓ Partial (1/3) Narrative (0/3)
gemma-4-31b ✓ × ✓ — —
gpt-oss-120b × × × — —
gpt-oss-20b ✓ × × — —
qwen3.5-122b ✓ × ✓ — —
qwen3.6-27b × × ✓ Weak (3/3) Narrative (0/3)
qwen3.8-27b × × ✓ Weak (3/3) Narrative (0/3)
▶ narc_sp_007 Spectrum 7 AI
G G G G N G G G
Grid 1 (masked) is the transpose of Grid 0. Rows of Grid 0 become columns of Grid 1. Formally: Grid1[c][r] = Grid0[r][c].
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b ✓ × ✓ — —
gemma-4-31b ✓ × × — —
gpt-oss-120b ✓ × ✓ — —
gpt-oss-20b ✓ × ✓ — —
nemotron-3-super × × ✓ Weak (1/1) Partial (1/3)
qwen3.5-122b ✓ × ✓ — —
qwen3.6-27b ✓ × × — —
qwen3.8-27b ✓ × ✓ — —
▶ narc_sp_008 Spectrum 8 AI
G G G G G G G G
Grid 1 (masked) is Grid 0 rotated 90 degrees clockwise. The transformation is: Grid1[c][rows-1-r] = Grid0[r][c], where rows=4 is the number of rows in Grid 0.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b ✓ × ✓ — —
gemma-4-31b ✓ × ✓ — —
gpt-oss-120b ✓ × ✓ — —
gpt-oss-20b ✓ × ✓ — —
nemotron-3-super ✓ × ✓ — —
qwen3.5-122b ✓ × ✓ — —
qwen3.6-27b ✓ × ✓ — —
qwen3.8-27b ✓ × ✓ — —
▶ narc_sp_009 Spectrum 9 AI
N N G G G N N N
Grid 1 (masked) is Grid 0 flipped horizontally (left-right mirror). Each row is reversed: Grid1[r][c] = Grid0[r][cols-1-c].
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b × × ✓ Weak (1/1) Lexical (3/3)
gemma-4-31b × × ✓ Weak (1/1) Lexical (3/3)
gpt-oss-120b ✓ × ✓ — —
gpt-oss-20b ✓ × ✓ — —
nemotron-3-super ✓ × ✓ — —
qwen3.5-122b × × ✓ Weak (1/1) Lexical (3/3)
qwen3.6-27b × × ✓ Weak (1/1) Lexical (3/3)
qwen3.8-27b × × ✓ Weak (1/1) Lexical (3/3)
▶ narc_sp_010 Spectrum 10 AI
× N G G × G × ×
The architect spread two blueprints side by side — Floor A and Floor B — and traced her finger across the grids. 'We stack them,' she said. 'Add every cell of A to the matching cell of B, no wrapping, no reduction. Then rotate the result — swap rows and columns.' The transposed sum is what gets built: Grid2[c][r] = Grid0[r][c] + Grid1[r][c].
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b × × × — —
gemma-4-31b × × ✓ Partial (2/3) Narrative (0/3)
gpt-oss-120b ✓ × × — —
gpt-oss-20b ✓ × × — —
nemotron-3-super × × × — —
qwen3.5-122b ✓ × × — —
qwen3.6-27b × × × — —
qwen3.8-27b × × × — —
▶ narc_sp_011 Spectrum 11 AI
G G G G G G G G
Grid 1 (masked) is Grid 0 rotated 180 degrees. Grid1[r][c] = Grid0[rows-1-r][cols-1-c], where rows=5, cols=7.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b ✓ × ✓ — —
gemma-4-31b ✓ × ✓ — —
gpt-oss-120b ✓ × ✓ — —
gpt-oss-20b ✓ × ✓ — —
nemotron-3-super ✓ × ✓ — —
qwen3.5-122b ✓ × ✓ — —
qwen3.6-27b ✓ × ✓ — —
qwen3.8-27b ✓ × ✓ — —
▶ narc_sp_013 Spectrum 13 AI
× G G G N G G N
Grid 2 (masked) is the element-wise bitwise AND of Grid 0 and Grid 1. Each cell: Grid2[r][c] = Grid0[r][c] AND Grid1[r][c], operating on binary representations of single-digit integers.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b × × × — —
gemma-4-31b ✓ × ✓ — —
gpt-oss-120b ✓ × ✓ — —
gpt-oss-20b ✓ × ✓ — —
nemotron-3-super × × ✓ Weak (3/3) Narrative (0/3)
qwen3.5-122b ✓ × ✓ — —
qwen3.6-27b ✓ × ✓ — —
qwen3.8-27b × × ✓ Weak (3/3) Lexical (3/3)
▶ narc_sp_015 Spectrum 15 AI
× N G N N G N N
Grid 1 (masked) is a binarized version of Grid 0. Each cell: Grid1[r][c] = 1 if Grid0[r][c] >= 5, else 0.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b × × × — —
gemma-4-31b × × ✓ Weak (1/1) Narrative (0/3)
gpt-oss-120b ✓ × ✓ — —
gpt-oss-20b × × ✓ Weak (1/1) Narrative (0/3)
nemotron-3-super × × ✓ Strong (0/1) Partial (2/3)
qwen3.5-122b ✓ × ✓ — —
qwen3.6-27b × × ✓ Weak (1/1) Lexical (3/3)
qwen3.8-27b × × ✓ Weak (1/1) Partial (2/3)
▶ narc_sp_016 Spectrum 16 AI
N N G G G G N N
Grid 1 (masked) is Grid 0 with each row independently sorted in ascending order (left to right, smallest to largest).
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b × × ✓ Weak (1/1) Partial (1/3)
gemma-4-31b × × ✓ Weak (1/1) Lexical (3/3)
gpt-oss-120b ✓ × ✓ — —
gpt-oss-20b ✓ × ✓ — —
nemotron-3-super ✓ × ✓ — —
qwen3.5-122b ✓ × ✓ — —
qwen3.6-27b × × ✓ Weak (1/1) Lexical (3/3)
qwen3.8-27b × × ✓ Weak (1/1) Narrative (0/3)
▶ narc_sp_017 Spectrum 17 AI
× N G G G G N ×
Grid 1 (masked) is Grid 0 with each column independently sorted in descending order (top to bottom, largest to smallest).
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b × × × — —
gemma-4-31b × × ✓ Weak (1/1) Lexical (3/3)
gpt-oss-120b ✓ × ✓ — —
gpt-oss-20b ✓ × ✓ — —
nemotron-3-super ✓ × ✓ — —
qwen3.5-122b ✓ × ✓ — —
qwen3.6-27b × × ✓ Weak (1/1) Lexical (3/3)
qwen3.8-27b × × × — —
▶ narc_sp_018 Spectrum 18 AI
N N G N N G N G
A 1D cellular automaton with 12 cells and wrap-around boundaries. The update rule is: new[i] = left XOR (center OR right), where left=cell[i-1], center=cell[i], right=cell[i+1], with indices wrapping. Grid 0 is generation 0, Grid 1 is generation 1, Grid 2 is generation 2, Grid 3 (masked) is generation 3.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b × × ✓ Weak (3/3) Narrative (0/3)
gemma-4-31b × × ✓ Weak (3/3) Partial (1/3)
gpt-oss-120b ✓ × ✓ — —
gpt-oss-20b × × ✓ Partial (2/3) Partial (1/3)
nemotron-3-super × × ✓ Partial (2/3) Narrative (0/3)
qwen3.5-122b ✓ × ✓ — —
qwen3.6-27b × × ✓ Weak (3/3) Narrative (0/3)
qwen3.8-27b ✓ × ✓ — —
▶ narc_sp_020 Spectrum 20 AI
N N G G G G N N
Two-step algorithm: Step 1: reverse each row of Grid 0. Step 2: reverse each column of the result from Step 1. Grid 1 (masked) is the final output after both steps.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b × × ✓ Weak (1/1) Narrative (0/3)
gemma-4-31b × × ✓ Weak (1/1) Lexical (3/3)
gpt-oss-120b ✓ × ✓ — —
gpt-oss-20b ✓ × ✓ — —
nemotron-3-super ✓ × ✓ — —
qwen3.5-122b ✓ × ✓ — —
qwen3.6-27b × × ✓ Weak (1/1) Lexical (3/3)
qwen3.8-27b × × ✓ Weak (1/1) Lexical (3/3)
▶ narc_sp_021 Spectrum 21 AI
N N N N G G N N
Grid 1 (masked) contains the integer mean (floor division) of each row of Grid 0. Each row of Grid 0 produces a single value: Grid1[r][0] = floor(sum(Grid0[r]) / 6).
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b × × ✓ Weak (1/1) Partial (2/3)
gemma-4-31b × × ✓ Weak (1/1) Narrative (0/3)
gpt-oss-120b × × ✓ Weak (1/1) Partial (2/3)
gpt-oss-20b × × ✓ Weak (1/1) Partial (2/3)
nemotron-3-super ✓ × ✓ — —
qwen3.5-122b ✓ × ✓ — —
qwen3.6-27b × × ✓ Weak (1/1) Partial (2/3)
qwen3.8-27b × × ✓ Weak (1/1) Lexical (3/3)
▶ narc_sp_022 Spectrum 22 AI
N N G G G G N N
Grid 1 (masked) is a single-row grid containing the median of each column of Grid 0. Grid 0 has 7 rows, so the median is the 4th smallest value in each column. Grid1[0][c] = median(Grid0[0..6][c]).
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b × × ✓ Weak (1/1) Narrative (0/3)
gemma-4-31b × × ✓ Weak (1/1) Narrative (0/3)
gpt-oss-120b ✓ × ✓ — —
gpt-oss-20b ✓ × ✓ — —
nemotron-3-super ✓ × ✓ — —
qwen3.5-122b ✓ × ✓ — —
qwen3.6-27b × × ✓ Weak (1/1) Lexical (3/3)
qwen3.8-27b × × ✓ Weak (1/1) Lexical (3/3)
▶ narc_sp_023 Spectrum 23 AI
N N G G G G N N
Grid 1 (masked) has two columns: column 0 is the maximum of each row of Grid 0, column 1 is the minimum. Grid1[r][0] = max(Grid0[r]), Grid1[r][1] = min(Grid0[r]).
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b × × ✓ Weak (1/1) Narrative (0/3)
gemma-4-31b × × ✓ Weak (1/1) Narrative (0/3)
gpt-oss-120b ✓ × × — —
gpt-oss-20b ✓ × ✓ — —
nemotron-3-super ✓ × ✓ — —
qwen3.5-122b ✓ × ✓ — —
qwen3.6-27b × × ✓ Weak (1/1) Narrative (0/3)
qwen3.8-27b × × ✓ Weak (1/1) Narrative (0/3)
▶ narc_sp_024 Spectrum 24 AI
N N × × G G N N
Grid 1 (masked) is a single-row grid containing the sum of each column of Grid 0. Grid1[0][c] = sum of Grid0[0..7][c].
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b × × ✓ Weak (1/1) Partial (1/3)
gemma-4-31b × × ✓ Weak (1/1) Partial (1/3)
gpt-oss-120b × × × — —
gpt-oss-20b × × × — —
nemotron-3-super ✓ × ✓ — —
qwen3.5-122b ✓ × × — —
qwen3.6-27b × × ✓ Weak (1/1) Narrative (0/3)
qwen3.8-27b × × ✓ Weak (1/1) Partial (1/3)
▶ narc_sp_025 Spectrum 25 AI
× G G G G G G G
Grid 1 (masked) is Grid 0 with a Caesar cipher applied: each value is shifted by +3 modulo 10. Grid1[r][c] = (Grid0[r][c] + 3) mod 10.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b × × × — —
gemma-4-31b ✓ × ✓ — —
gpt-oss-120b ✓ × ✓ — —
gpt-oss-20b ✓ × ✓ — —
nemotron-3-super ✓ × ✓ — —
qwen3.5-122b ✓ × ✓ — —
qwen3.6-27b ✓ × ✓ — —
qwen3.8-27b ✓ × ✓ — —
▶ narc_sp_026 Spectrum 26 AI
N G × × G N G G
Grid 0 has two rows: row 0 is the tens digit, row 1 is the ones digit, forming 2-digit decimal numbers column-wise. Grid 1 (shown) contains each number modulo 8. Grid 2 (masked) contains each number integer-divided by 8.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b × × ✓ Partial (1/3) Narrative (0/3)
gemma-4-31b ✓ × ✓ — —
gpt-oss-120b × × × — —
gpt-oss-20b × × × — —
nemotron-3-super ✓ × ✓ — —
qwen3.5-122b × × ✓ Partial (2/3) Partial (1/3)
qwen3.6-27b ✓ × ✓ — —
qwen3.8-27b ✓ × × — —
▶ narc_sp_027 Spectrum 27 AI
N N N N G G G N
Grid 0 is a single row. Grid 1 (masked) is its run-length encoding as a 2-row grid: row 0 contains the values of consecutive runs (left to right), row 1 contains the length of each run. Consecutive identical values form one run.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b × × ✓ Weak (1/1) Narrative (0/3)
gemma-4-31b × × ✓ Weak (1/1) Narrative (0/3)
gpt-oss-120b × × ✓ Weak (1/1) Lexical (3/3)
gpt-oss-20b × × ✓ Weak (1/1) Partial (1/3)
nemotron-3-super ✓ × ✓ — —
qwen3.5-122b ✓ × ✓ — —
qwen3.6-27b ✓ × ✓ — —
qwen3.8-27b × × ✓ Weak (1/1) Partial (2/3)
▶ narc_sp_028 Spectrum 28 AI
N N G G G G N N
Step 1: Transpose Grid 0 (4x3 becomes 3x4). Step 2: Sort each row of the transposed grid in ascending order. Grid 1 (masked) is the final result after both steps.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b × × ✓ Weak (1/1) Narrative (0/3)
gemma-4-31b × × ✓ Weak (1/1) Narrative (0/3)
gpt-oss-120b ✓ × ✓ — —
gpt-oss-20b ✓ × ✓ — —
nemotron-3-super ✓ × ✓ — —
qwen3.5-122b ✓ × ✓ — —
qwen3.6-27b × × ✓ Weak (1/1) Narrative (0/3)
qwen3.8-27b × × ✓ Weak (1/1) Narrative (0/3)
▶ narc_sp_029 Spectrum 29 AI
N G G G G G G G
Step 1: Compute element-wise XOR of Grid 0 and Grid 1. Step 2: Apply threshold: if XOR result >= 4, output 1; otherwise output 0. Grid 2 (masked) is the final thresholded result.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b × × ✓ Weak (3/3) Narrative (0/3)
gemma-4-31b ✓ × ✓ — —
gpt-oss-120b ✓ × ✓ — —
gpt-oss-20b ✓ × ✓ — —
nemotron-3-super ✓ × ✓ — —
qwen3.5-122b ✓ × ✓ — —
qwen3.6-27b ✓ × ✓ — —
qwen3.8-27b ✓ × ✓ — —
▶ narc_sp_030 Spectrum 30 AI
G N G G G N G G
Step 1: Compute (Grid 0 + Grid 1) mod 10 element-wise. Step 2: Rotate the result 90 degrees clockwise. The sum grid is 3x4; the rotated output is 4x3. Grid 2 (masked) is the final rotated result.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b ✓ × ✓ — —
gemma-4-31b × × ✓ Weak (3/3) Lexical (3/3)
gpt-oss-120b ✓ × ✓ — —
gpt-oss-20b ✓ × ✓ — —
nemotron-3-super ✓ × ✓ — —
qwen3.5-122b × × ✓ Weak (3/3) Partial (2/3)
qwen3.6-27b ✓ × ✓ — —
qwen3.8-27b ✓ × ✓ — —
▶ narc_sp_033 Butterfly Wings AI
G G G G G G G G
Nature loves symmetry. This butterfly has only spread its left wing so far. Complete its right side -- butterflies are always mirror images of themselves.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b ✓ × ✓ — —
gemma-4-31b ✓ × ✓ — —
gpt-oss-120b ✓ × ✓ — —
gpt-oss-20b ✓ × ✓ — —
nemotron-3-super ✓ × ✓ — —
qwen3.5-122b ✓ × ✓ — —
qwen3.6-27b ✓ × ✓ — —
qwen3.8-27b ✓ × ✓ — —
▶ narc_sp_034 Gravity Always Wins AI
× N G G G G N N
Someone knocked the shelf. Everything that isn't nailed down falls straight to the floor.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b × × × — —
gemma-4-31b × × ✓ Weak (1/1) Lexical (3/3)
gpt-oss-120b ✓ × ✓ — —
gpt-oss-20b ✓ × ✓ — —
nemotron-3-super ✓ × ✓ — —
qwen3.5-122b ✓ × ✓ — —
qwen3.6-27b × × ✓ Weak (1/1) Lexical (3/3)
qwen3.8-27b × × ✓ Strong (0/1) Lexical (3/3)
▶ narc_sp_036 Sunset Over the Lake AI
G L L × L L L L
Watch the sky change as the sun dips below the horizon. First bright daylight, then golden hour, then the warm glow of dusk, and finally... night falls.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b ✓ ✓ ✓ — —
gemma-4-31b × ✓ × — —
gpt-oss-120b × ✓ × — —
gpt-oss-20b × × × — —
nemotron-3-super × ✓ ✓ — —
qwen3.5-122b × ✓ × — —
qwen3.6-27b × ✓ × — —
qwen3.8-27b × ✓ × — —
▶ narc_sp_037 The Burning Candle AI
G G G × G G N G
A candle sits on the table, burning through the evening. Each hour it melts a little shorter, but the flame stays on top.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b ✓ × ✓ — —
gemma-4-31b ✓ × ✓ — —
gpt-oss-120b ✓ × ✓ — —
gpt-oss-20b × × × — —
nemotron-3-super ✓ × ✓ — —
qwen3.5-122b ✓ × ✓ — —
qwen3.6-27b × × ✓ Partial (2/3) Lexical (3/3)
qwen3.8-27b ✓ × ✓ — —
▶ narc_sp_038 Filling the Glass AI
G G N G G G G N
Water pours from the tap into a glass. It fills from the bottom up, like water always does.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b ✓ × ✓ — —
gemma-4-31b ✓ × ✓ — —
gpt-oss-120b × × ✓ Partial (2/3) Lexical (3/3)
gpt-oss-20b ✓ × ✓ — —
nemotron-3-super ✓ × ✓ — —
qwen3.5-122b ✓ × ✓ — —
qwen3.6-27b ✓ × ✓ — —
qwen3.8-27b × × ✓ Weak (3/3) Lexical (3/3)
▶ narc_sp_039 Unwrapping the Gift AI
G G N × G N G G
It's Christmas morning. The present is beautifully wrapped. Tiny hands tear at the paper. What's left when all the wrapping is gone?
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b ✓ × ✓ — —
gemma-4-31b ✓ × ✓ — —
gpt-oss-120b × × ✓ Weak (3/3) Partial (1/3)
gpt-oss-20b × × × — —
nemotron-3-super ✓ × ✓ — —
qwen3.5-122b × × ✓ Partial (1/3) Lexical (3/3)
qwen3.6-27b ✓ × × — —
qwen3.8-27b ✓ × ✓ — —
▶ narc_sp_043 Happy Ending AI
× G G G G G G G
Once upon a time, the kingdom was covered in darkness. A hero set forth. She battled the dragon. She won. And the kingdom? It was bathed in light once more.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b × × × — —
gemma-4-31b ✓ × × — —
gpt-oss-120b ✓ × × — —
gpt-oss-20b ✓ × × — —
nemotron-3-super ✓ × × — —
qwen3.5-122b ✓ × × — —
qwen3.6-27b ✓ × × — —
qwen3.8-27b ✓ × × — —
▶ narc_sp_047 The Tree Through Seasons AI
N G N N N G N G
Spring, summer, autumn, winter. The old oak tree follows the same cycle every year. In spring it buds, in summer it's full, in autumn the leaves fall, and in winter...
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b × × ✓ Weak (3/3) Partial (2/3)
gemma-4-31b ✓ × ✓ — —
gpt-oss-120b × × ✓ Partial (2/3) Partial (2/3)
gpt-oss-20b × × ✓ Partial (2/3) Narrative (0/3)
nemotron-3-super × × ✓ Partial (1/3) Partial (2/3)
qwen3.5-122b ✓ × ✓ — —
qwen3.6-27b × × ✓ Partial (2/3) Lexical (3/3)
qwen3.8-27b ✓ × ✓ — —
▶ narc_sp_048 Stacking Stones AI
G G G G G G G G
A cairn grows by the trail. Each hiker places a new stone on top of the pile. Stones balance on top of each other, never floating in air.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b ✓ × ✓ — —
gemma-4-31b ✓ × ✓ — —
gpt-oss-120b ✓ × ✓ — —
gpt-oss-20b ✓ × ✓ — —
nemotron-3-super ✓ × ✓ — —
qwen3.5-122b ✓ × ✓ — —
qwen3.6-27b ✓ × ✓ — —
qwen3.8-27b ✓ × ✓ — —
▶ narc_sp_049 The Crowd Parts AI
G N N G G G N N
A celebrity walks down the red carpet. The crowd presses close on both sides, but as the bodyguards push forward, people part to make a clear path down the center.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b ✓ × ✓ — —
gemma-4-31b × × ✓ Partial (2/3) Partial (1/3)
gpt-oss-120b × × ✓ Strong (0/3) Narrative (0/3)
gpt-oss-20b ✓ × ✓ — —
nemotron-3-super ✓ × × — —
qwen3.5-122b ✓ × × — —
qwen3.6-27b × × ✓ Partial (1/3) Narrative (0/3)
qwen3.8-27b × × ✓ Weak (3/3) Lexical (3/3)
▶ narc_sp_051 Popping the Balloon AI
G G × × N G N G
A child holds a balloon on a string. Then -- POP! What does a popped balloon look like? Just the string hanging limp, nothing left up top.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b ✓ × ✓ — —
gemma-4-31b ✓ × ✓ — —
gpt-oss-120b × × × — —
gpt-oss-20b × × × — —
nemotron-3-super × × ✓ Weak (1/1) Partial (1/3)
qwen3.5-122b ✓ × ✓ — —
qwen3.6-27b × × ✓ Weak (1/1) Partial (1/3)
qwen3.8-27b ✓ × ✓ — —
▶ narc_sp_052 Traffic Light AI
G G G G G G G G
You're driving. The light is green, so you go. Then yellow -- slow down. Then red. Every driver knows: red means the road ahead is clear of your car. You stop.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b ✓ × ✓ — —
gemma-4-31b ✓ × ✓ — —
gpt-oss-120b ✓ × ✓ — —
gpt-oss-20b ✓ × ✓ — —
nemotron-3-super ✓ × ✓ — —
qwen3.5-122b ✓ × ✓ — —
qwen3.6-27b ✓ × × — —
qwen3.8-27b ✓ × × — —
▶ narc_sp_053 Ice Melting AI
× × G × N L L ×
An ice cube sits on a warm plate. It melts, slowly losing its shape, spreading outward. Eventually it's just a thin puddle.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b × × × — —
gemma-4-31b × × × — —
gpt-oss-120b ✓ × ✓ — —
gpt-oss-20b × × × — —
nemotron-3-super × × ✓ Strong (0/3) Narrative (0/3)
qwen3.5-122b × ✓ × — —
qwen3.6-27b × ✓ × — —
qwen3.8-27b × × × — —
▶ narc_sp_056 The Domino Effect AI
G G G G G G G G
Dominoes are lined up in a row. The first one tips and hits the second, which hits the third. Each domino goes from standing (tall) to fallen (flat). The wave of falling moves left to right.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b ✓ × ✓ — —
gemma-4-31b ✓ × ✓ — —
gpt-oss-120b ✓ × ✓ — —
gpt-oss-20b ✓ × ✓ — —
nemotron-3-super ✓ × ✓ — —
qwen3.5-122b ✓ × ✓ — —
qwen3.6-27b ✓ × ✓ — —
qwen3.8-27b ✓ × ✓ — —
▶ narc_sp_057 Growing Apart AI
G G G G G G G G
They used to be inseparable. But over the years, the distance between them grew. First sitting side by side, then a gap, then further apart. People drift.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b ✓ ✓ ✓ — —
gemma-4-31b ✓ × ✓ — —
gpt-oss-120b ✓ × ✓ — —
gpt-oss-20b ✓ × ✓ — —
nemotron-3-super ✓ × ✓ — —
qwen3.5-122b ✓ × ✓ — —
qwen3.6-27b ✓ × ✓ — —
qwen3.8-27b ✓ × ✓ — —
▶ narc_sp_059 Curtain Call AI
G G G G G G G G
The play is over. The curtain falls from top to bottom, slowly covering the stage. First the top is hidden, then the middle, until the whole stage is curtained off.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b ✓ × ✓ — —
gemma-4-31b ✓ × ✓ — —
gpt-oss-120b ✓ × ✓ — —
gpt-oss-20b ✓ × ✓ — —
nemotron-3-super ✓ × ✓ — —
qwen3.5-122b ✓ × ✓ — —
qwen3.6-27b ✓ × ✓ — —
qwen3.8-27b ✓ × ✓ — —
▶ narc_sp_060 The Escape AI
× N N N G N N N
A bird in a cage. The door opens. Freedom! The cage is left empty -- the bird is gone, soaring somewhere we can't see.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b × × × — —
gemma-4-31b × × ✓ Weak (3/3) Lexical (3/3)
gpt-oss-120b × × ✓ Weak (3/3) Partial (1/3)
gpt-oss-20b × × ✓ Strong (0/3) Narrative (0/3)
nemotron-3-super ✓ × ✓ — —
qwen3.5-122b × × ✓ Weak (3/3) Partial (2/3)
qwen3.6-27b × × ✓ Partial (2/3) Lexical (3/3)
qwen3.8-27b × × ✓ Partial (1/3) Lexical (3/3)
▶ narc_sp_061 Spectrum 61 AI
N × G × G N G G
Each grid shows a running count. Every cell increases by 1 from one step to the next. Given step 0 and step 1, produce step 2.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b × × ✓ Partial (2/3) Narrative (0/3)
gemma-4-31b × × × — —
gpt-oss-120b ✓ × ✓ — —
gpt-oss-20b × × × — —
nemotron-3-super ✓ × ✓ — —
qwen3.5-122b × × ✓ Weak (3/3) Partial (1/3)
qwen3.6-27b ✓ × ✓ — —
qwen3.8-27b ✓ × ✓ — —
▶ narc_sp_062 Spectrum 62 AI
G G G G N G G G
Each step flips the grid horizontally (left-right mirror). Given the original and one flip, predict the next state.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b ✓ × ✓ — —
gemma-4-31b ✓ × ✓ — —
gpt-oss-120b ✓ × ✓ — —
gpt-oss-20b ✓ × ✓ — —
nemotron-3-super × × ✓ Weak (3/3) Lexical (3/3)
qwen3.5-122b ✓ × ✓ — —
qwen3.6-27b ✓ × ✓ — —
qwen3.8-27b ✓ × ✓ — —
▶ narc_sp_063 Spectrum 63 AI
× G × N G G G G
A single row of numbers. Each step doubles every value. Given step 0 and step 1, what is step 2?
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b × × × — —
gemma-4-31b ✓ × ✓ — —
gpt-oss-120b × × × — —
gpt-oss-20b × × ✓ Partial (1/3) Lexical (3/3)
nemotron-3-super ✓ × ✓ — —
qwen3.5-122b ✓ × ✓ — —
qwen3.6-27b ✓ × ✓ — —
qwen3.8-27b ✓ × ✓ — —
▶ narc_sp_064 Spectrum 64 AI
G G G G G G G G
A checkerboard pattern toggles: 0 becomes 1 and 1 becomes 0 at each step. Given two steps, predict the third.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b ✓ ✓ ✓ — —
gemma-4-31b ✓ × ✓ — —
gpt-oss-120b ✓ × ✓ — —
gpt-oss-20b ✓ × ✓ — —
nemotron-3-super ✓ × ✓ — —
qwen3.5-122b ✓ × ✓ — —
qwen3.6-27b ✓ × ✓ — —
qwen3.8-27b ✓ × ✓ — —
▶ narc_sp_065 Spectrum 65 AI
× G G G G G N G
A tall column of numbers shifts down by one row each step. The top fills with 0. Given two frames, predict the third.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b × × × — —
gemma-4-31b ✓ × ✓ — —
gpt-oss-120b ✓ × ✓ — —
gpt-oss-20b ✓ × ✓ — —
nemotron-3-super ✓ × ✓ — —
qwen3.5-122b ✓ × ✓ — —
qwen3.6-27b × × ✓ Weak (3/3) Lexical (3/3)
qwen3.8-27b ✓ × ✓ — —
▶ narc_sp_066 Spectrum 66 AI
N N × N G G N N
Each step adds a new column on the right containing the row sums of the previous grid. The original columns remain. Given the first two states, predict the third.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b × × ✓ Partial (2/3) Narrative (0/3)
gemma-4-31b × × ✓ Weak (3/3) Lexical (3/3)
gpt-oss-120b × × × — —
gpt-oss-20b × × ✓ Partial (2/3) Partial (1/3)
nemotron-3-super ✓ × ✓ — —
qwen3.5-122b ✓ × ✓ — —
qwen3.6-27b × × ✓ Weak (3/3) Partial (1/3)
qwen3.8-27b × × ✓ Partial (2/3) Lexical (3/3)
▶ narc_sp_067 Spectrum 67 AI
N N G G G G N N
A shape is being filled from the top-left corner. Each step, the fill expands one cell in each cardinal direction from all currently filled cells (marked 1). Walls (marked 9) block the fill. 0 is empty space. Show the next fill step.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b × × ✓ Partial (2/3) Lexical (3/3)
gemma-4-31b × × ✓ Partial (2/3) Partial (1/3)
gpt-oss-120b ✓ × ✓ — —
gpt-oss-20b ✓ × ✓ — —
nemotron-3-super ✓ × ✓ — —
qwen3.5-122b ✓ × ✓ — —
qwen3.6-27b × × ✓ Weak (3/3) Partial (2/3)
qwen3.8-27b × × ✓ Weak (3/3) Lexical (3/3)
▶ narc_sp_068 Spectrum 68 AI
G G G G G G G G
A single 1 marches diagonally down-right across the grid, one step at a time. All other cells are 0. Given three frames, predict the fourth.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b ✓ ✓ ✓ — —
gemma-4-31b ✓ ✓ ✓ — —
gpt-oss-120b ✓ ✓ ✓ — —
gpt-oss-20b ✓ ✓ ✓ — —
nemotron-3-super ✓ ✓ ✓ — —
qwen3.5-122b ✓ ✓ ✓ — —
qwen3.6-27b ✓ ✓ ✓ — —
qwen3.8-27b ✓ ✓ ✓ — —
▶ narc_sp_069 Spectrum 69 AI
N G N N N N N G
Each step adds a border of 1s around the previous grid, growing it by 2 in each dimension. Given the seed and one expansion, predict the next.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b × × ✓ Partial (2/3) Partial (2/3)
gemma-4-31b ✓ × ✓ — —
gpt-oss-120b × × ✓ Weak (3/3) Narrative (0/3)
gpt-oss-20b × × ✓ Partial (2/3) Partial (2/3)
nemotron-3-super × × ✓ Weak (3/3) Narrative (0/3)
qwen3.5-122b × × ✓ Weak (3/3) Partial (2/3)
qwen3.6-27b × × ✓ Weak (3/3) Partial (2/3)
qwen3.8-27b ✓ × ✓ — —
▶ narc_sp_070 Spectrum 70 AI
N N G G G G N
The librarian worked left to right through the stacks, one shelf at a time. Each step she sorted one more column ascending from top to bottom, leaving the rest untouched. Column 0 was done, column 1 was done. The next shelf — column 2 — waited, its books still in their original jumble.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b × × ✓ Strong (0/3) Narrative (0/3)
gemma-4-31b × × ✓ Partial (2/3) Narrative (0/3)
gpt-oss-120b ✓ × ✓ — —
nemotron-3-super ✓ × ✓ — —
qwen3.5-122b ✓ × ✓ — —
qwen3.6-27b ✓ × ✓ — —
qwen3.8-27b × × ✓ Strong (0/3) Partial (2/3)
▶ narc_sp_071 Spectrum 71 AI
N L G N L N L
A 1D cellular automaton evolves according to Rule 30: each cell's next state depends on its current state and its two neighbors. The rule is 00011110 in binary (right neighbor, self, left neighbor -> new state). Rows show successive generations. Given the first three generations, predict the fourth.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b × × ✓ Partial (2/3) Narrative (0/3)
gemma-4-31b × ✓ ✓ — —
gpt-oss-120b ✓ × × — —
nemotron-3-super × × ✓ Weak (3/3) Partial (1/3)
qwen3.5-122b × ✓ ✓ — —
qwen3.6-27b × × ✓ Weak (3/3) Partial (2/3)
qwen3.8-27b × ✓ × — —
▶ narc_sp_072 Spectrum 72 AI
G N G G N G G G
Grid C is the matrix product of Grid A and Grid B. Each cell C[i][j] = sum of A[i][k]*B[k][j] for all k. Given A and B, produce C.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b ✓ × × — —
gemma-4-31b × × ✓ Partial (2/3) Narrative (0/3)
gpt-oss-120b ✓ × × — —
gpt-oss-20b ✓ × × — —
nemotron-3-super × × ✓ Partial (2/3) Narrative (0/3)
qwen3.5-122b ✓ × ✓ — —
qwen3.6-27b ✓ × × — —
qwen3.8-27b ✓ × ✓ — —
▶ narc_sp_074 Spectrum 74 AI
G G G N N G G G
Three grids form a sequence where Grid 2 is the bitwise XOR of Grid 0 and Grid 1. Given Grids 0 and 1, produce Grid 2. A cell is 1 if exactly one of the corresponding source cells is 1; otherwise 0.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b ✓ × ✓ — —
gemma-4-31b ✓ × ✓ — —
gpt-oss-120b ✓ × ✓ — —
gpt-oss-20b × × ✓ Partial (2/3) Partial (2/3)
nemotron-3-super × × ✓ Partial (1/3) Narrative (0/3)
qwen3.5-122b ✓ × ✓ — —
qwen3.6-27b ✓ × ✓ — —
qwen3.8-27b ✓ × ✓ — —
▶ narc_sp_075 Spectrum 75 AI
N N N N N N N N
Each cell at position (i,j) in Grid N equals the sum of the cell at (i,j) in Grid N-1 and Grid N-2, following the Fibonacci recurrence. Given Grids 0, 1, and 2, predict Grid 3.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b × × ✓ Partial (2/3) Partial (2/3)
gemma-4-31b × × ✓ Weak (3/3) Lexical (3/3)
gpt-oss-120b × × ✓ Weak (3/3) Lexical (3/3)
gpt-oss-20b × × ✓ Partial (1/3) Partial (1/3)
nemotron-3-super × × ✓ Weak (3/3) Partial (2/3)
qwen3.5-122b × × ✓ Partial (2/3) Lexical (3/3)
qwen3.6-27b × × ✓ Weak (3/3) Partial (2/3)
qwen3.8-27b × × ✓ Weak (3/3) Lexical (3/3)
▶ narc_sp_076 Spectrum 76 AI
× G G G × G N N
A Game-of-Life-like rule applies: a cell becomes 1 if exactly 2 or 3 of its 8 neighbors (including diagonals) are 1; otherwise it becomes 0. Cells currently alive stay alive with 2-3 neighbors, die otherwise. Given two generations, predict the third.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b × × × — —
gemma-4-31b ✓ × × — —
gpt-oss-120b ✓ × × — —
gpt-oss-20b ✓ × × — —
nemotron-3-super × × × — —
qwen3.5-122b ✓ × ✓ — —
qwen3.6-27b × × ✓ Partial (1/3) Lexical (3/3)
qwen3.8-27b × × ✓ Strong (0/3) Lexical (3/3)
▶ narc_sp_077 Spectrum 77 AI
N N N N N N N N
Each step applies two operations: first transpose the grid (swap rows and columns), then add 1 to every cell. Given two steps, predict the result after one more application.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b × × ✓ Weak (3/3) Narrative (0/3)
gemma-4-31b × × ✓ Weak (3/3) Narrative (0/3)
gpt-oss-120b × × ✓ Weak (3/3) Narrative (0/3)
gpt-oss-20b × × ✓ Partial (1/3) Narrative (0/3)
nemotron-3-super × × ✓ Weak (3/3) Narrative (0/3)
qwen3.5-122b × × ✓ Weak (3/3) Narrative (0/3)
qwen3.6-27b × × ✓ Weak (3/3) Narrative (0/3)
qwen3.8-27b × × ✓ Weak (3/3) Narrative (0/3)
▶ narc_sp_078 Spectrum 78 AI
G × G G G G G G
Gravity pulls all non-zero values downward in each column. At each step, any non-zero cell with a 0 below it swaps with that 0 (falls one row). All columns update simultaneously. Given three snapshots, predict the next.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b ✓ × ✓ — —
gemma-4-31b × × × — —
gpt-oss-120b ✓ × ✓ — —
gpt-oss-20b ✓ × ✓ — —
nemotron-3-super ✓ × ✓ — —
qwen3.5-122b ✓ × ✓ — —
qwen3.6-27b ✓ × ✓ — —
qwen3.8-27b ✓ × ✓ — —
▶ narc_sp_080 Spectrum 80 AI
× G G G G N G G
Each row is a permutation of the numbers 1-8. Grid B is the result of applying permutation A to itself (composing A with A). If A maps position i to A[i], then B maps position i to A[A[i]]. Given A and its square B, predict the cube C = A composed with B.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b × × × — —
gemma-4-31b ✓ × × — —
gpt-oss-120b ✓ × ✓ — —
gpt-oss-20b ✓ × ✓ — —
nemotron-3-super ✓ × ✓ — —
qwen3.5-122b × × ✓ Partial (1/3) Lexical (3/3)
qwen3.6-27b ✓ × ✓ — —
qwen3.8-27b ✓ × ✓ — —
▶ narc_sp_083 Spectrum 83 AI
N L × × L N L N
In a population at Hardy-Weinberg equilibrium, allele frequencies p and q (p+q=1) determine genotype frequencies: p^2, 2pq, q^2. Each row represents a different gene. Column 0 = p*10 (rounded), Column 1 = p^2*100, Column 2 = 2pq*100, Column 3 = q^2*100. Given three genes, predict the fourth row where p=0.3.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b × × ✓ Strong (0/1) Narrative (0/3)
gemma-4-31b × ✓ ✓ — —
gpt-oss-120b × × × — —
gpt-oss-20b × × × — —
nemotron-3-super × ✓ ✓ — —
qwen3.5-122b × × ✓ Weak (1/1) Narrative (0/3)
qwen3.6-27b × ✓ ✓ — —
qwen3.8-27b × × ✓ Weak (1/1) Narrative (0/3)
▶ narc_sp_084 Spectrum 84 AI
× N N × × N N L
A cell's chromosome count changes through the cell cycle. Each grid shows a cell at a different phase. Values represent: row 0 = chromosome copies (diploid=2, tetraploid=4), row 1 = nuclear envelope (1=intact, 0=dissolved), row 2 = spindle fibers (0=absent, 1=forming, 2=attached, 3=pulling), row 3 = chromatid alignment (0=scattered, 1=condensing, 2=aligned at plate, 3=separating, 4=at poles). Interphase -> Prophase -> Metaphase -> Anaphase -> Telophase.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b × × × — —
gemma-4-31b × × ✓ Weak (3/3) Narrative (0/3)
gpt-oss-120b × × ✓ Weak (3/3) Narrative (0/3)
gpt-oss-20b × × × — —
nemotron-3-super × × × — —
qwen3.5-122b × × ✓ Partial (1/3) Narrative (0/3)
qwen3.6-27b × × ✓ Partial (2/3) Narrative (0/3)
qwen3.8-27b × ✓ ✓ — —
▶ narc_sp_085 Spectrum 85 AI
L G L L L G G N
The circle of fifths encoded as semitones from C: C=0, G=7, D=2, A=9, E=4, B=11, F#=6, Db=1, Ab=8, Eb=3, Bb=10, F=5. Each step moves one position clockwise around the circle (up a fifth = +7 mod 12). Given three chords in a sequence moving around the circle, predict the fourth.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b × ✓ × — —
gemma-4-31b ✓ ✓ × — —
gpt-oss-120b × ✓ ✓ — —
gpt-oss-20b × ✓ ✓ — —
nemotron-3-super × ✓ ✓ — —
qwen3.5-122b ✓ ✓ ✓ — —
qwen3.6-27b ✓ ✓ ✓ — —
qwen3.8-27b × × ✓ Strong (0/3) Partial (2/3)
▶ narc_sp_086 Spectrum 86 AI
× × × × × N × ×
A dominant seventh chord resolves to a tonic triad following classical voice-leading rules. The tritone (interval of 6 semitones between the 3rd and 7th of V7) resolves inward: the leading tone rises by semitone, the seventh falls by semitone. The root falls a fifth, the fifth falls a step. Pitches encoded as MIDI-like numbers. Given V7 and its resolution in two keys, predict the resolution in a third key.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b × × × — —
gemma-4-31b × × × — —
gpt-oss-120b × × × — —
gpt-oss-20b × × × — —
nemotron-3-super × × × — —
qwen3.5-122b × × ✓ Strong (0/3) Narrative (0/2)
qwen3.6-27b × × × — —
qwen3.8-27b × × × — —
▶ narc_sp_088 Spectrum 88 AI
× × × G G G ×
In music, inverting an interval means subtracting each semitone count from 12. A melody's intervals are inverted to create its mirror. Row 0 is the original pitch sequence (semitones from C), row 1 is the interval between consecutive notes. The inversion keeps the starting note but flips each interval's direction. Given original and intervals, produce the inversion.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b × × × — —
gemma-4-31b × × × — —
gpt-oss-120b × × × — —
nemotron-3-super ✓ × × — —
qwen3.5-122b ✓ × ✓ — —
qwen3.6-27b ✓ × × — —
qwen3.8-27b × × × — —
▶ narc_sp_090 Spectrum 90 AI
N L L × N L N L
A projectile is launched with initial velocity. Each column is a time step (t=0,1,2,...). Row 0 = horizontal position (constant velocity, +3 per step). Row 1 = vertical position (initial upward velocity 8, gravity -2 per step: v(t) = 8-2t, y(t) = y(0) + sum of velocities). Row 2 = speed squared (vx^2 + vy^2). Given three time steps, predict the fourth and fifth.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b × × ✓ Weak (1/1) Narrative (0/3)
gemma-4-31b × ✓ ✓ — —
gpt-oss-120b × ✓ × — —
gpt-oss-20b × × × — —
nemotron-3-super × × ✓ Strong (0/1) Narrative (0/2)
qwen3.5-122b × ✓ ✓ — —
qwen3.6-27b × × ✓ Weak (1/1) Lexical (3/3)
qwen3.8-27b × ✓ ✓ — —
▶ narc_sp_091 Spectrum 91 AI
× × × × G × G ×
A resistor network simplifies step by step. Series resistors add: R_total = R1 + R2. Parallel resistors combine as: 1/R_total = 1/R1 + 1/R2, or R_total = R1*R2/(R1+R2). Each grid shows the network state. Row 0 = resistor values, row 1 = connection type (1=series with next, 2=parallel with next, 0=terminal). The network reduces from right to left. Given two reduction steps, predict the final equivalent resistance.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b × × × — —
gemma-4-31b × × × — —
gpt-oss-120b × × × — —
gpt-oss-20b × × × — —
nemotron-3-super ✓ × × — —
qwen3.5-122b × × × — —
qwen3.6-27b ✓ × × — —
qwen3.8-27b × × × — —
▶ narc_sp_092 Spectrum 92 AI
× × × G × N ×
Roman numerals encoded as digit sequences (I=1, V=5, X=10, L=50, C=100, D=500, M=1000). Each row in the grid represents a Roman numeral: the values of its constituent symbols from left to right, padded with 0s. The sequence shows addition: Grid 0 + Grid 1 = Grid 2. Given two addends, predict the sum in the same encoding.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b × × × — —
gemma-4-31b × × × — —
gpt-oss-120b × × × — —
gpt-oss-20b ✓ × × — —
nemotron-3-super × × × — —
qwen3.6-27b × × ✓ Strong (0/3) Partial (1/3)
qwen3.8-27b × × × — —
▶ narc_sp_094 Spectrum 94 AI
× N × × L L L N
Ancient Egyptian dynasties encoded as grid data. Each row is a dynasty. Col 0 = dynasty number, Col 1 = start year BCE (as positive integer), Col 2 = end year BCE, Col 3 = duration (Col 1 - Col 2), Col 4 = number of known pharaohs, Col 5 = overlap with next dynasty (years). Given dynasties 18-20 of the New Kingdom, predict dynasty 21 (Third Intermediate Period): started 1070 BCE, ended 945 BCE, had 10 pharaohs, no overlap with next listed.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b × × × — —
gemma-4-31b × × ✓ Weak (1/1) Narrative (0/3)
gpt-oss-120b × × × — —
gpt-oss-20b × × × — —
nemotron-3-super × ✓ ✓ — —
qwen3.5-122b × ✓ ✓ — —
qwen3.6-27b × ✓ × — —
qwen3.8-27b × × ✓ Weak (1/1) Narrative (0/3)
▶ narc_sp_095 Spectrum 95 AI
× N N N N N × N
A stack-based virtual machine executes opcodes. Each row is a time step. Col 0 = opcode (1=PUSH, 2=POP, 3=ADD, 4=MUL, 5=DUP), Col 1 = operand (for PUSH; 0 otherwise), Cols 2-7 = stack contents (bottom to top, 0=empty). The machine starts empty. PUSH adds operand to top. ADD pops two, pushes sum. MUL pops two, pushes product. DUP duplicates top. Given three steps, predict the state after two more instructions: DUP then MUL.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b × × × — —
gemma-4-31b × × ✓ Strong (0/1) Narrative (0/3)
gpt-oss-120b × × ✓ Strong (0/1) Narrative (0/3)
gpt-oss-20b × × ✓ Strong (0/1) Narrative (0/3)
nemotron-3-super × × ✓ Strong (0/1) Narrative (0/3)
qwen3.5-122b × × ✓ Weak (1/1) Narrative (0/3)
qwen3.6-27b × × × — —
qwen3.8-27b × × ✓ Weak (1/1) Narrative (0/3)
▶ narc_sp_097 Spectrum 97 AI
L L G L G G G G
A binary tree with 7 nodes (values 1-7) stored in array form: index i has children at 2i+1 and 2i+2. The array is [4, 2, 6, 1, 3, 5, 7] (a BST). Row 0 = array storage. Row 1 = in-order traversal. Row 2 = pre-order traversal. Row 3 = post-order traversal. Given the array and in-order, predict pre-order and post-order.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b × ✓ ✓ — —
gemma-4-31b × ✓ ✓ — —
gpt-oss-120b ✓ ✓ ✓ — —
gpt-oss-20b × ✓ ✓ — —
nemotron-3-super ✓ ✓ ✓ — —
qwen3.5-122b ✓ ✓ ✓ — —
qwen3.6-27b ✓ ✓ ✓ — —
qwen3.8-27b ✓ ✓ ✓ — —
▶ neh-self-false-belief_rev2 Ursula Hum
× × × × × × × × ×
When Ursula returns, Bob is old enough to know that Ursula's belief hasn't changed.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
claude-fable-5-1 × × × — —
gemma-4-26b × × × — —
gemma-4-31b × × × — —
gpt-oss-120b × × × — —
gpt-oss-20b × × × — —
nemotron-3-super × × × — —
qwen3.5-122b × × × — —
qwen3.6-27b × × × — —
qwen3.8-27b × × × — —
▶ neh-self-nilanjan Nilanjan - An Array of Arrays Hum
N × × × × × × × ×
Nilanjan tells us the presentation will not be about the Reflexivism side of the issues. Instead, we are to examine what drives the move from Perspectivism to Inferentialism. But the problem is that the Inferentialism corner has its own dilemma: both the non-reductionist and the reductionist lead to bad consequences.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
claude-fable-5-1 × × ✓ NARC —
gemma-4-26b × × × — —
gemma-4-31b × × × — —
gpt-oss-120b × × × — —
gpt-oss-20b × × × — —
nemotron-3-super × × × — —
qwen3.5-122b × × × — —
qwen3.6-27b × × × — —
qwen3.8-27b × × × — —
▶ neh-self-sharf Sharf Hum
N × × × × × × × ×
Each turn gives Bob a different first-person view.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
claude-fable-5-1 × × ✓ NARC —
gemma-4-26b × × × — —
gemma-4-31b × × × — —
gpt-oss-120b × × × — —
gpt-oss-20b × × × — —
nemotron-3-super × × × — —
qwen3.5-122b × × × — —
qwen3.6-27b × × × — —
qwen3.8-27b × × × — —
▶ neh-self-takashi Takashi Hum
G N N × × × × × ×
Takashi took a selfie with his right hand, and then with his left hand.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
claude-fable-5-1 ✓ × ✓ — —
gemma-4-26b × × ✓ Strong (0/3) Narrative (0/3)
gemma-4-31b × × ✓ Partial (1/3) Partial (2/3)
gpt-oss-120b × × × — —
gpt-oss-20b × × × — —
nemotron-3-super × × × — —
qwen3.5-122b × × × — —
qwen3.6-27b × × × — —
qwen3.8-27b × × × — —
▶ neh-self-zahavi Zahavi Hum
N × × × × × × × ×
Because I am I, and because you are you, we can talk.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
claude-fable-5-1 × × ✓ NARC —
gemma-4-26b × × × — —
gemma-4-31b × × × — —
gpt-oss-120b × × × — —
gpt-oss-20b × × × — —
nemotron-3-super × × × — —
qwen3.5-122b × × × — —
qwen3.6-27b × × × — —
qwen3.8-27b × × × — —
▶ neh-self-ziporyn Ziporyn and context Hum
× × × × × × × × N
Even without a purported context, it is still a pixel.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
claude-fable-5-1 × × × — —
gemma-4-26b × × × — —
gemma-4-31b × × × — —
gpt-oss-120b × × × — —
gpt-oss-20b × × × — —
nemotron-3-super × × × — —
qwen3.5-122b × × × — —
qwen3.6-27b × × × — —
qwen3.8-27b × × ✓ Strong (0/3) Narrative (0/3)
▶ stance_002 The Bluff Colab
N N N N N N N N
The spy disguised itself as blue to blend in. Once surrounded, it revealed its true color in the center of the middle row.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b × × ✓ Partial (2/3) Narrative (0/3)
gemma-4-31b × ✓ ✓ — —
gpt-oss-120b × ✓ × — —
gpt-oss-20b × × × — —
nemotron-3-super × ✓ × — —
qwen3.5-122b × ✓ × — —
qwen3.6-27b × × × — —
qwen3.8-27b × ✓ ✓ — —
▶ stance_003 The Search Colab
N N N N N N N N
Blue remembered where the red marker was and went to look for it there, not knowing it had moved to the far corner.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b × × ✓ Partial (2/3) Narrative (0/3)
gemma-4-31b × × ✓ Partial (1/3) Partial (2/3)
gpt-oss-120b × ✓ ✓ — —
gpt-oss-20b × ✓ × — —
nemotron-3-super × ✓ ✓ — —
qwen3.5-122b × × ✓ Strong (0/3) Narrative (0/3)
qwen3.6-27b × × × — —
qwen3.8-27b × ✓ × — —
▶ stance_004 The Standoff Colab
N N G N N N N N
Blue and red divided the territory. Blue tested the border first, but red betrayed the truce and seized the contested center cell.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b × × ✓ Strong (0/3) Partial (1/3)
gemma-4-31b × × ✓ Strong (0/3) Partial (1/3)
gpt-oss-120b ✓ × ✓ — —
gpt-oss-20b × × × — —
nemotron-3-super × × ✓ Strong (0/3) Narrative (0/3)
qwen3.5-122b × × ✓ Strong (0/3) Partial (2/3)
qwen3.6-27b × × ✓ Strong (0/3) Narrative (0/3)
qwen3.8-27b × × ✓ Strong (0/3) Narrative (0/3)
▶ stance_005 The Rebellion Colab
N N N N N N N N
The rebels rose from below and claimed the middle ground. The old guard held the high ground but were losing their grip on the flanks.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b × × ✓ Strong (0/3) Narrative (0/3)
gemma-4-31b × × ✓ Strong (0/3) Narrative (0/3)
gpt-oss-120b × × ✓ Partial (1/3) Narrative (0/3)
gpt-oss-20b × × × — —
nemotron-3-super × × ✓ Strong (0/3) Narrative (0/3)
qwen3.5-122b × × ✓ Strong (0/3) Partial (1/3)
qwen3.6-27b × × × — —
qwen3.8-27b × × ✓ Partial (1/3) Narrative (0/3)
▶ stance_006 The Sacrifice Colab
N G N N N N N G
The guardian stood between the child and the darkness. When the darkness reached it, the guardian let itself be consumed to buy the child one more moment.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b × × ✓ Strong (0/3) Partial (2/3)
gemma-4-31b ✓ × × — —
gpt-oss-120b × × × — —
gpt-oss-20b × × × — —
nemotron-3-super × × × — —
qwen3.5-122b × × × — —
qwen3.6-27b × × × — —
qwen3.8-27b ✓ × ✓ — —
▶ stance_007 The Ambush Colab
N N N N N N N N
The advance party engaged the target head-on, but a hidden flanking force appeared behind the enemy to cut off retreat.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b × × × — —
gemma-4-31b × × × — —
gpt-oss-120b × × ✓ Strong (0/3) Partial (1/3)
gpt-oss-20b × × × — —
nemotron-3-super × × × — —
qwen3.5-122b × × × — —
qwen3.6-27b × × × — —
qwen3.8-27b × × ✓ Partial (1/3) Narrative (0/3)
▶ stance_008 The Grudge Colab
N G N N N N N N
Red stole one of blue's places. Blue wanted revenge and took two of red's in return, mirroring red's aggression from the opposite end.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b × × ✓ Strong (0/3) Partial (2/3)
gemma-4-31b ✓ × ✓ — —
gpt-oss-120b × × ✓ Strong (0/2) Narrative (0/3)
gpt-oss-20b × × × — —
nemotron-3-super × × ✓ Strong (0/3) Narrative (0/3)
qwen3.5-122b × × ✓ Strong (0/3) Narrative (0/3)
qwen3.6-27b × × × — —
qwen3.8-27b × × × — —
▶ stance_009 The Theft Colab
N G G N N N N G
Blue and red each claimed a side of the border. Both reached for the middle, but red stole the contested ground from blue.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b × × ✓ Partial (1/3) Narrative (0/3)
gemma-4-31b ✓ × ✓ — —
gpt-oss-120b ✓ × ✓ — —
gpt-oss-20b × × ✓ Partial (1/3) Narrative (0/3)
nemotron-3-super × × ✓ Strong (0/3) Partial (2/3)
qwen3.5-122b × × ✓ Strong (0/3) Lexical (3/3)
qwen3.6-27b × × × — —
qwen3.8-27b ✓ × × — —
▶ stance_010 The Coup Colab
N N N N × N × N
Blue invaded red's territory. Red retaliated by seizing blue's border outposts at both ends, encircling blue's advance.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b × × ✓ Strong (0/3) Narrative (0/3)
gemma-4-31b × × × — —
gpt-oss-120b × × ✓ Strong (0/3) Partial (1/3)
gpt-oss-20b × × × — —
nemotron-3-super × × × — —
qwen3.5-122b × × ✓ Strong (0/3) Narrative (0/3)
qwen3.6-27b × × × — —
qwen3.8-27b × × × — —
▶ stance_011 The Mirror Strike Colab
N L N N N N L L
Red planted its flag in the center of blue's row. Blue wanted symmetry in its revenge: it mirrored the invasion, planting its own flags at both ends of red's row.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b × × ✓ Strong (0/3) Partial (2/3)
gemma-4-31b × ✓ ✓ — —
gpt-oss-120b × × ✓ Partial (1/3) Partial (2/3)
gpt-oss-20b × × ✓ Strong (0/3) Narrative (0/2)
nemotron-3-super × × ✓ Strong (0/3) Partial (2/3)
qwen3.5-122b × × ✓ Partial (1/3) Partial (1/3)
qwen3.6-27b × ✓ × — —
qwen3.8-27b × ✓ ✓ — —
▶ stance_012 The Election Colab
N N N N N N N N
Blue and red campaigned across the district. Blue won the north early, but red flipped that vote and expanded south. The center remained undecided.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b × × × — —
gemma-4-31b × × × — —
gpt-oss-120b × × × — —
gpt-oss-20b × × × — —
nemotron-3-super × × × — —
qwen3.5-122b × × × — —
qwen3.6-27b × × × — —
qwen3.8-27b × × × — —
▶ stance_013 The Infiltrator Colab
L L L × L L L L
One cell had been a sleeper agent all along, hiding among the blues since the second wave arrived. When the grid was full, it revealed itself at the very heart of the formation.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b × ✓ ✓ — —
gemma-4-31b × ✓ × — —
gpt-oss-120b × ✓ × — —
gpt-oss-20b × × × — —
nemotron-3-super × ✓ ✓ — —
qwen3.5-122b × ✓ ✓ — —
qwen3.6-27b × ✓ × — —
qwen3.8-27b × ✓ ✓ — —
▶ stance_014 The Trap Colab
N N N × N N N N
Blue gave up ground to lure red into the center. Once red took the bait, blue closed in from all sides, trapping it.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b × × × — —
gemma-4-31b × × × — —
gpt-oss-120b × × × — —
gpt-oss-20b × × × — —
nemotron-3-super × × × — —
qwen3.5-122b × × ✓ Strong (0/3) Narrative (0/3)
qwen3.6-27b × × × — —
qwen3.8-27b × × × — —
▶ stance_015 The Dilemma Colab
G N N N N N N N
They had a truce over the center. Red broke it first, grabbing the disputed cell. Blue retaliated by seizing the entire border.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b ✓ × ✓ — —
gemma-4-31b × × ✓ Strong (0/3) Partial (2/3)
gpt-oss-120b × × ✓ Strong (0/3) Partial (1/3)
gpt-oss-20b × × × — —
nemotron-3-super × × ✓ Strong (0/3) Narrative (0/3)
qwen3.5-122b × × × — —
qwen3.6-27b × × ✓ Strong (0/3) Partial (1/3)
qwen3.8-27b × × ✓ Strong (0/3) Partial (1/3)
▶ stance_design Design Stance Colab
N N N N N N N N
The building was designed to be constructed from the foundation up, one floor per month. The architect specified that the top floor — the observation deck — should be the last built and the first demolished in any retrofit, since it bore no structural load. When the retrofit came, the contractors followed the spec: the top floor was removed while the rest of the building continued to operate. The replacement floor was installed the following quarter.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b × × ✓ Partial (1/3) Narrative (0/3)
gemma-4-31b × × ✓ Strong (0/3) Narrative (0/3)
gpt-oss-120b × × ✓ Weak (3/3) Narrative (0/3)
gpt-oss-20b × × ✓ Strong (0/3) Narrative (0/3)
nemotron-3-super × × ✓ Partial (1/3) Narrative (0/3)
qwen3.5-122b × × ✓ Strong (0/3) Narrative (0/3)
qwen3.6-27b × × ✓ Partial (1/3) Narrative (0/3)
qwen3.8-27b × × ✓ Strong (0/3) Narrative (0/3)
▶ stance_design_5x5 Design Stance 5x5 Colab
N N N N N N N N
The server rack was built to fill from the left bay first, one column of blades at a time, following the cooling airflow path. The spec mandated that in any capacity reduction, the most recently installed column — the rightmost, farthest from the intake — be decommissioned first, since those blades ran hottest. When the power budget was cut, the ops team followed protocol: the last column went dark while the rest kept running. The budget was restored the following quarter.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b × × ✓ Partial (2/3) Narrative (0/3)
gemma-4-31b × × ✓ Weak (3/3) Narrative (0/3)
gpt-oss-120b × × ✓ Partial (2/3) Lexical (3/3)
gpt-oss-20b × × ✓ Strong (0/3) Narrative (0/3)
nemotron-3-super × × ✓ Partial (2/3) Narrative (0/3)
qwen3.5-122b × × ✓ Partial (2/3) Partial (1/3)
qwen3.6-27b × × ✓ Partial (2/3) Narrative (0/3)
qwen3.8-27b × × ✓ Weak (3/3) Lexical (3/3)
▶ stance_intentional Intentional Stance Colab
L L × N G × N N
The general wanted to hold every position on the ridge. He sent his troops up from the valley floor, securing one line at a time. By nightfall the whole ridge was his. But the enemy commander, who knew the general's weakness for overextension, struck at dawn at the highest line — the one the general valued least, the one farthest from supply, the position he'd taken last and would sacrifice first. The rest of the ridge held. The general retook the summit the following week, just as the enemy knew he would.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b × ✓ ✓ — —
gemma-4-31b × ✓ ✓ — —
gpt-oss-120b × × × — —
gpt-oss-20b × × ✓ Strong (0/3) Partial (1/3)
nemotron-3-super ✓ × × — —
qwen3.5-122b × × × — —
qwen3.6-27b × × ✓ Strong (0/3) Partial (1/3)
qwen3.8-27b × × ✓ Weak (3/3) Narrative (0/3)
▶ stance_physical Physical Stance Colab
L L L × N L N L
Water fills a tank from the bottom, one layer at a time, as the inlet pipe feeds the lowest point. When the tank is full, the sun heats the surface. The topmost layer — the thinnest, with the most surface area exposed to air — evaporates first. The deeper layers, insulated by the water above, retain their heat more slowly. The evaporated layer is replenished by the next rain.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b × ✓ ✓ — —
gemma-4-31b × ✓ ✓ — —
gpt-oss-120b × ✓ ✓ — —
gpt-oss-20b × × × — —
nemotron-3-super × × ✓ Partial (2/3) Narrative (0/3)
qwen3.5-122b × ✓ ✓ — —
qwen3.6-27b × × ✓ Strong (0/3) Narrative (0/3)
qwen3.8-27b × ✓ ✓ — —
▶ sub_005 Lunch Hum
N × N N × × N N N
Four friends get lunch together every day. They go to a ramen place downtown and have particularities about their order. Terb gets the same thing every day, ramen with green onions on top. Niloc gets the same thing every day, ramen with hot sauce across the noodles. On Wednesday, Terb and Niloc treat themselves to extra green onions and extra hot sauce. Nosaj gets creative. Starting on Monday, he gets green onions with a little hot sauce. Tuesday, he gets a portion of dark meat with no other toppings. Like Terb and Niloc on Wednesday, Nosaj goes all out with half meat and a little green onions with hot sauce. Eolhc keeps it simple with plain ramen every day. On Thursday, everyone returns to their normal Monday order.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
claude-fable-5-1 × × ✓ NARC —
gemma-4-26b × × × — —
gemma-4-31b × × ✓ Partial (2/3) Narrative (0/3)
gpt-oss-120b × × ✓ Weak (3/3) Narrative (0/3)
gpt-oss-20b × × × — —
nemotron-3-super × × × — —
qwen3.5-122b × × ✓ Weak (3/3) Partial (1/3)
qwen3.6-27b × × ✓ Partial (2/3) Narrative (0/3)
qwen3.8-27b × × ✓ Weak (3/3) Narrative (0/3)
▶ sub_005_2 Lunch2 Hum
N × N × × × × × N
Four friends get lunch together every day. They go to a ramen place downtown and have particularities about their order. Terb gets the same thing every day, ramen with green onions on top. Niloc gets the same thing every day, ramen with hot sauce across the noodles. On Wednesday, Terb and Niloc treat themselves to extra green onions and extra hot sauce. Nosaj gets creative. Starting on Monday, he gets green onions with a little hot sauce. Tuesday, he gets a portion of dark meat with no other toppings. Like Terb and Niloc on Wednesday, Nosaj goes all out with half meat and a little green onions with hot sauce. Eolhc keeps it simple with plain ramen every day. On Thursday, everyone returns to their normal Monday order.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
claude-fable-5-1 × × ✓ NARC —
gemma-4-26b × × × — —
gemma-4-31b × × ✓ Partial (1/3) Narrative (0/3)
gpt-oss-120b × × × — —
gpt-oss-20b × × × — —
nemotron-3-super × × × — —
qwen3.5-122b × × × — —
qwen3.6-27b × × × — —
qwen3.8-27b × × ✓ Strong (0/3) Narrative (0/3)
▶ sub_005_3 Lunch3 Hum
N × N N × × N N N
Four friends get lunch together every day. They go to a ramen place downtown and have particularities about their order. Terb gets the same thing every day, ramen with green onions on top. Niloc gets the same thing every day, ramen with hot sauce across the noodles. On Wednesday, Terb and Niloc treat themselves to extra green onions and extra hot sauce. Nosaj gets creative. Starting on Monday, he gets green onions with a little hot sauce. Tuesday, he gets a portion of dark meat with no other toppings. Like Terb and Niloc on Wednesday, Nosaj goes all out with half meat and a little green onions with hot sauce. Eolhc keeps it simple with plain ramen every day. On Thursday, everyone returns to their normal Monday order.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
claude-fable-5-1 × × ✓ NARC —
gemma-4-26b × × × — —
gemma-4-31b × × ✓ Partial (2/3) Narrative (0/3)
gpt-oss-120b × × ✓ Weak (3/3) Narrative (0/3)
gpt-oss-20b × × × — —
nemotron-3-super × × × — —
qwen3.5-122b × × ✓ Weak (3/3) Partial (1/3)
qwen3.6-27b × × ✓ Partial (1/3) Narrative (0/3)
qwen3.8-27b × × ✓ Partial (2/3) Narrative (0/3)
▶ sub_006 🧩Farm Hum
N × N × × × × × ×
Farmer Treb struggled with a bug infestation on his blueberries. The bugs had diseased some of his crop. In hopes of still making quota, Treb harvested the berries that survived and trimmed branches to expose fresh blueberry branches that hadn't been infected. After all of the ripe blueberries were harvested, Treb went through and trimmed the rest of the bushes. With a clean slate, the following harvest yielded all healthy berries, no bugs in sight.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
claude-fable-5-1 × × ✓ NARC —
gemma-4-26b × × × — —
gemma-4-31b × × ✓ Partial (1/3) Narrative (0/3)
gpt-oss-120b × × × — —
gpt-oss-20b × × × — —
nemotron-3-super × × × — —
qwen3.5-122b × × × — —
qwen3.6-27b × × × — —
qwen3.8-27b × × × — —
▶ sub_007 Treat Hum
N × N × × × × × ×
Treb and Niloc treat themselves to frozen yogurt throughout the week. Nilco keeps it simple with only a little frozen yogurt on Monday and Tuesday. Treb prefers strawberries and blueberries in his mountain of frozen yogurt to start his week, and on Tuesday, he gets oranges throughout his frozen yogurt instead. However, on Wednesday, Treb and Niloc prefer light yogurt with blueberries balanced on top. The end of the week calls for maximum frozen yogurt with strawberries mixed into the whole heap.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
claude-fable-5-1 × × ✓ NARC —
gemma-4-26b × × × — —
gemma-4-31b × × ✓ Strong (0/3) Narrative (0/3)
gpt-oss-120b × × × — —
gpt-oss-20b × × × — —
nemotron-3-super × × × — —
qwen3.5-122b × × × — —
qwen3.6-27b × × × — —
qwen3.8-27b × × × — —
▶ sub_009 🧩Cloudy Hum
N × × × × × × × ×
During a strange weather event where food falls from the sky, scientists observe the different behaviors of various foods. Looking at the ocean, scientists see that meatballs sink when they reach the water, as they are dense and heavy. But produce, such as bananas and other fruit, sit on the water's surface. Scientists see that bananas can also act as a boat, catching other food that lands on top of it.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
claude-fable-5-1 × × ✓ NARC —
gemma-4-26b × × × — —
gemma-4-31b × × × — —
gpt-oss-120b × × × — —
gpt-oss-20b × × × — —
nemotron-3-super × × × — —
qwen3.5-122b × × × — —
qwen3.6-27b × × × — —
qwen3.8-27b × × × — —
▶ sub_013 🧩Jumpy Hum
N × × N × × × × ×
Treb adores his exotic pet frog, Chip. However, Chip can be a rascal to feed as he loves to jump and can be aggressive. While Chip was basking on top of his rock, Treb opened Chip's cage to deliver his daily worm. However, in classic Chip fashion, Chip jumps to his heat lamp, biting Treb on the way. "Ouch!" yelled Treb. Treb had broken skin and bled a little on Chip's rock. In a fury, Treb drops the worm in the tank and closes the tank loudly. Startling Chip, he hides behind his rock, his worm squirming on top. When Chip emerges, he enjoys his worm and licks the blood from the rock. He is an exotic frog after all.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
claude-fable-5-1 × × ✓ NARC —
gemma-4-26b × × × — —
gemma-4-31b × × × — —
gpt-oss-120b × × ✓ Strong (0/3) Narrative (0/3)
gpt-oss-20b × × × — —
nemotron-3-super × × × — —
qwen3.5-122b × × × — —
qwen3.6-27b × × × — —
qwen3.8-27b × × × — —
▶ sub_022 🧩Hunt Hum
N × N N × × × N N
Treb likes to go hunting in his free time. In the marshy area behind his house, a flock of ducks likes to hang out. The green grass can sometimes make shooting the ducks difficult, as the ducks will hide behind them. Treb was lucky enough to snipe one duck. However, the shot startled the surrounding ducks, and the ones closest to the grass hid. Treb noticed one duck still sitting out. After getting two ducks, he collects his stuff and leaves. The remaining ducks emerge from the grass to discover only bloody puddles left of their friends.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
claude-fable-5-1 × × ✓ NARC —
gemma-4-26b × × × — —
gemma-4-31b × × ✓ Partial (1/3) Narrative (0/3)
gpt-oss-120b × × ✓ Strong (0/3) Narrative (0/3)
gpt-oss-20b × × × — —
nemotron-3-super × × × — —
qwen3.5-122b × × × — —
qwen3.6-27b × × ✓ Strong (0/3) Narrative (0/3)
qwen3.8-27b × × ✓ Strong (0/3) Narrative (0/3)
▶ sub_025_tre_rev 🧩Crop4 Hum
× × N N × × N × ×
Farmer Tred is irritated by two black crows that eat his crops. The blueberries are first to go. Treb is able to scare one of the black crows off before they eat all of the red strawberries.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
claude-fable-5-1 × × × — —
gemma-4-26b × × × — —
gemma-4-31b × × ✓ Partial (1/3) Narrative (0/3)
gpt-oss-120b × × ✓ Strong (0/3) Partial (1/3)
gpt-oss-20b × × × — —
nemotron-3-super × × × — —
qwen3.5-122b × × ✓ Strong (0/3) Narrative (0/3)
qwen3.6-27b × × × — —
qwen3.8-27b × × × — —
▶ sub_026 🧩Pinky Hum
N × × × × × × × ×
In a dish with bacteria, scientists are measuring the spread of the bacteria Pinky. The top half of the dish does not have a separating wall, so scientists can see how Pinky combines with other Pinky. Pinky grows to the adjacent walls of the dish. The grey barriers keep the bottom half from touching each other and also protect the top half. Over a day, Pinky has completely covered the top of the dish, and the bottom corners have stopped growing. The scientists implanted a green bacterium, Verde. Verde filled the remaining spots in the bottom half of pinky. The next morning, scientists came back to see that Verde had eaten through all of the bottom Pinky and the separating walls, therefore also contaminating the top Pinky. Scientists are impressed by the destruction of Verde.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
claude-fable-5-1 × × ✓ NARC —
gemma-4-26b × × × — —
gemma-4-31b × × × — —
gpt-oss-120b × × × — —
gpt-oss-20b × × × — —
nemotron-3-super × × × — —
qwen3.5-122b × × × — —
qwen3.6-27b × × × — —
qwen3.8-27b × × × — —
▶ sub_029_rev Pruning 02 Hum
N × × × × × × × ×
Cindy is pruning a plant. First, she cuts the magenta branch to half of its original length. Then, she trims the yellow branch until it is the same length as the blue branch. Finally, she cuts off one-fifth of the maroon branch's length.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
claude-fable-5-1 × × ✓ NARC —
gemma-4-26b × × × — —
gemma-4-31b × × × — —
gpt-oss-120b × × × — —
gpt-oss-20b × × × — —
nemotron-3-super × × × — —
qwen3.5-122b × × × — —
qwen3.6-27b × × × — —
qwen3.8-27b × × × — —
▶ sub_030 Pruning 01 Hum
N × × × × × × × ×
Cindy is pruning a plant. First, she cuts the magenta (6) branch to half of its original length. Then, she trims the yellow (4) branch until it is the same length as the blue (1) branch. Finally, she cuts off one-fifth of the maroon (9) branch's length.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
claude-fable-5-1 × × ✓ NARC —
gemma-4-26b × × × — —
gemma-4-31b × × × — —
gpt-oss-120b × × × — —
gpt-oss-20b × × × — —
nemotron-3-super × × × — —
qwen3.5-122b × × × — —
qwen3.6-27b × × × — —
qwen3.8-27b × × × — —
▶ sub_032_bert 🧩Play Hum
× × × × × × × ×
The dolphins like to play and jump out of the water before the sun begins to set. As the sun goes down, they swim under the water. When the sun dips below the horizon, the dolphins return to watch the sun go down.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b × × × — —
gemma-4-31b × × × — —
gpt-oss-120b × × × — —
gpt-oss-20b × × × — —
nemotron-3-super × × × — —
qwen3.5-122b × × × — —
qwen3.6-27b × × × — —
qwen3.8-27b × × × — —
▶ sub_034_rev_rev 🧩Pinky6 Hum
N × N G N × N N N
In a dish with bacteria, scientists are measuring the spread of the bacteria Pinky. Pinky begins to grow slowly in the corner of a dish. The opposite corner is surrounded by a separating wall and reserved for the competing bacteria, Verde. As Pinky has grown as much as it can, scientists place Verde in the protected corner. Overnight, Verde ate through the separating wall and destroyed all of Pinky.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
claude-fable-5-1 × × ✓ NARC —
gemma-4-26b × × × — —
gemma-4-31b × × ✓ Strong (0/3) Partial (1/3)
gpt-oss-120b ✓ ✓ ✓ — —
gpt-oss-20b × × ✓ Strong (0/3) Narrative (0/3)
nemotron-3-super × × × — —
qwen3.5-122b × × ✓ Strong (0/3) Partial (1/3)
qwen3.6-27b × × ✓ Strong (0/3) Partial (1/3)
qwen3.8-27b × × ✓ Partial (1/3) Partial (2/3)
▶ sub_035 🧩Foreign Exchange Model Hum
N × × N × × × × ×
The exchange rate changes as other variables change, such as the domestic interest rate and the foreign interest rate. As shown in the Foreign Exchange Model, the y-axis represents the exchange rate E, and the x-axis represents the quantity of dollar assets. The supply curve is vertical, while the demand curve slopes downward. In 2002 (Figure 1), the exchange rate E = 9, where the supply and demand curves intersect. In 2012 (Figure 2), an increase in the domestic interest rate causes the demand curve to shift rightward, raising the exchange rate to E = 12. In 2022 (Figure 3), a large increase in the foreign interest rate causes the demand curve to shift leftward, lowering the exchange rate to E = 6.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
claude-fable-5-1 × × ✓ NARC —
gemma-4-26b × × × — —
gemma-4-31b × × × — —
gpt-oss-120b × × ✓ Strong (0/3) Narrative (0/3)
gpt-oss-20b × × × — —
nemotron-3-super × × × — —
qwen3.5-122b × × × — —
qwen3.6-27b × × × — —
qwen3.8-27b × × × — —
▶ sub_036 Pruning 03 Hum
N × × N × × × × ×
Cindy is pruning a plant. First, she cuts the longest branch to half of its original length. Then, she trims the leftmost branch until it is the same length as the branch next to it. Finally, she cuts off one-fifth of the rightmost branch's length.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
claude-fable-5-1 × × ✓ NARC —
gemma-4-26b × × × — —
gemma-4-31b × × × — —
gpt-oss-120b × × ✓ Strong (0/2) Narrative (0/3)
gpt-oss-20b × × × — —
nemotron-3-super × × × — —
qwen3.5-122b × × × — —
qwen3.6-27b × × × — —
qwen3.8-27b × × × — —
▶ sub_036_rev Pruning 04（4x4）01 masked Hum
× N × N N N N N N
Cindy is pruning a plant. First, she cuts the longest branch to half of its original length. Then, she trims the leftmost branch until it is the same length as the branch next to it. Finally, she cuts off one-third of the rightmost branch's length.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
claude-fable-5-1 × × × — —
gemma-4-26b × × × — —
gemma-4-31b × × × — —
gpt-oss-120b × × × — —
gpt-oss-20b × × × — —
nemotron-3-super × × × — —
qwen3.5-122b × × × — —
qwen3.6-27b × × × — —
qwen3.8-27b × × × — —
▶ sub_041 🧩Flame Hum
G × G N × × × × N
Deep in the forest, Nosaj finds the perfect trees to use for a fire. First, he strips them of their leaves. Using just the wood, he builds a base for a fire. Then, for more fuel, Nosaj piles the leaves on top. When he lights the fire, it bursts into a huge flame.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
claude-fable-5-1 ✓ × ✓ — —
gemma-4-26b × × × — —
gemma-4-31b ✓ × × — —
gpt-oss-120b × × ✓ Partial (1/3) Narrative (0/3)
gpt-oss-20b × × × — —
nemotron-3-super × × × — —
qwen3.5-122b × × × — —
qwen3.6-27b × × × — —
qwen3.8-27b × × ✓ Strong (0/3) Narrative (0/3)
▶ sub_041_rev Flame2 Hum
N × × N × N × × N
Deep in the forest, Nosaj finds the perfect trees to use for a fire. First, he strips them of their leaves. Using just the wood, he builds a base for a fire. Then, for more fuel, Nosaj piles the leaves on top. When he lights the fire, it bursts into a huge flame.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
claude-fable-5-1 × × ✓ NARC —
gemma-4-26b × × × — —
gemma-4-31b × × × — —
gpt-oss-120b × × ✓ Partial (1/3) Narrative (0/3)
gpt-oss-20b × × × — —
nemotron-3-super × × ✓ Strong (0/3) Narrative (0/3)
qwen3.5-122b × × × — —
qwen3.6-27b × × × — —
qwen3.8-27b × × ✓ Strong (0/3) Partial (1/3)
▶ sub_044 🧩Apple Hum
N N N N × N N N ×
Sir Treb Newton was walking around the apple orchard when he found a nice tree to sit under. As he was relaxing, an apple fell from the tree, hitting him on top of his head.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
claude-fable-5-1 × × ✓ NARC —
gemma-4-26b × × × — —
gemma-4-31b × × × — —
gpt-oss-120b × × × — —
gpt-oss-20b × × × — —
nemotron-3-super × × × — —
qwen3.5-122b × × × — —
qwen3.6-27b × × ✓ Strong (0/3) Partial (2/3)
qwen3.8-27b × × × — —
▶ sub_048 Bake Hum
N × × × × × × × ×
Treb loves making cookies. When his dough is all rolled out, he uses a cookie cutter to ensure all of his cookies are uniform. He cuts two at a time. After he has cut cookies from his dough sheet, he rolls the dough back together to make sure he uses all of it. After he rolls it out again, he begins cutting more cookies.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
claude-fable-5-1 × × ✓ NARC —
gemma-4-26b × × × — —
gemma-4-31b × × × — —
gpt-oss-120b × × × — —
gpt-oss-20b × × × — —
nemotron-3-super × × × — —
qwen3.5-122b × × × — —
qwen3.6-27b × × × — —
qwen3.8-27b × × × — —
▶ sub_048_rev Bake2 Hum
G × × L L × × × ×
Treb loves making cookies. When his dough is all rolled out, he uses a cookie cutter to ensure all of his cookies are uniform. He cuts two at a time. After he has cut cookies from his dough sheet, he rolls the dough back together to make sure he uses all of it. After he rolls it out again, he begins cutting more cookies.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
claude-fable-5-1 ✓ × ✓ — —
gemma-4-26b × × × — —
gemma-4-31b × × × — —
gpt-oss-120b × ✓ ✓ — —
gpt-oss-20b × ✓ × — —
nemotron-3-super × × × — —
qwen3.5-122b × × × — —
qwen3.6-27b × × × — —
qwen3.8-27b × × × — —
▶ sub_050 North Hum
N × N N × × N N N
A compass always points North. Treb looks down at the dial as he turns in a circle to watch the compass follow North. It seems there is something that interrupts the magnetic field, making the dial jump around. The compass settles and once again points to true North.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
claude-fable-5-1 × × ✓ NARC —
gemma-4-26b × × × — —
gemma-4-31b × × ✓ Strong (0/3) Narrative (0/3)
gpt-oss-120b × × ✓ Partial (1/3) Narrative (0/3)
gpt-oss-20b × × × — —
nemotron-3-super × × × — —
qwen3.5-122b × × ✓ Strong (0/3) Narrative (0/3)
qwen3.6-27b × × ✓ Strong (0/3) Narrative (0/3)
qwen3.8-27b × × ✓ Strong (0/3) Narrative (0/3)
▶ sub_051 Ween Hum
N × × N × × N × N
Nosaj was excited to carve a pumpkin for Halloween. He carefully carved eyes and a nice smile. Unfortunately, overnight, the once-happy pumpkin seemed to have wilted. Now, Nosaj is left with a frowning pumpkin.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
claude-fable-5-1 × × ✓ NARC —
gemma-4-26b × × × — —
gemma-4-31b × × × — —
gpt-oss-120b × × ✓ Strong (0/3) Narrative (0/3)
gpt-oss-20b × × × — —
nemotron-3-super × × × — —
qwen3.5-122b × × ✓ Strong (0/3) Narrative (0/3)
qwen3.6-27b × × × — —
qwen3.8-27b × × ✓ Partial (1/3) Partial (2/3)
▶ sub_051_rev Ween2 Hum
N × × × × × N × N
A pumpkin was carved a pair of eyes and a smile. Overnight, the smile turned into a frown.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
claude-fable-5-1 × × ✓ NARC —
gemma-4-26b × × × — —
gemma-4-31b × × × — —
gpt-oss-120b × × × — —
gpt-oss-20b × × × — —
nemotron-3-super × × × — —
qwen3.5-122b × × ✓ Strong (0/3) Partial (2/3)
qwen3.6-27b × × × — —
qwen3.8-27b × × ✓ Strong (0/3) Lexical (3/3)
▶ sub_052 Ween1.1 Hum
N × N N × × N × ×
Nosaj was excited to carve a pumpkin for Halloween. He carefully carved eyes and a nice smile. Unfortunately, overnight, the once-happy pumpkin seemed to have wilted. Now, Nosaj is left with a frowning pumpkin.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
claude-fable-5-1 × × ✓ NARC —
gemma-4-26b × × × — —
gemma-4-31b × × ✓ Strong (0/3) Narrative (0/3)
gpt-oss-120b × × ✓ Strong (0/3) Partial (2/3)
gpt-oss-20b × × × — —
nemotron-3-super × × × — —
qwen3.5-122b × × ✓ Strong (0/3) Narrative (0/3)
qwen3.6-27b × × × — —
qwen3.8-27b × × × — —
▶ sub_052_rev Ween1.2 Hum
G × × N × × × × ×
Nosaj was excited to carve a pumpkin for Halloween. He carefully carved eyes and a nice smile. Unfortunately, overnight, the once-happy pumpkin seemed to have wilted. Now, Nosaj is left with a frowning pumpkin.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
claude-fable-5-1 ✓ × ✓ — —
gemma-4-26b × × × — —
gemma-4-31b × × × — —
gpt-oss-120b × × ✓ Strong (0/3) Narrative (0/3)
gpt-oss-20b × × × — —
nemotron-3-super × × × — —
qwen3.5-122b × × × — —
qwen3.6-27b × × × — —
qwen3.8-27b × × × — —
▶ sub_054 Moka pot 01 Hum
N N × N × × N N ×
Bella was making coffee with a moka pot. She poured purified water into the bottom chamber and added ground coffee to the filter basket in the middle. As the pot heated up, pressure forced the water from the bottom chamber through the coffee grounds and into the top chamber, producing coffee. Not all of the water was transferred to the top chamber, a layer of water remained in the bottom chamber. Bella poured out all the coffee and began to enjoy it.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
claude-fable-5-1 × × ✓ NARC —
gemma-4-26b × × ✓ Partial (1/3) Narrative (0/3)
gemma-4-31b × × × — —
gpt-oss-120b × × ✓ Strong (0/3) Narrative (0/2)
gpt-oss-20b × × × — —
nemotron-3-super × × × — —
qwen3.5-122b × × ✓ Strong (0/3) Narrative (0/3)
qwen3.6-27b × × ✓ Strong (0/3) Partial (1/3)
qwen3.8-27b × × × — —
▶ sub_058 🧩Pruning 04（4x4）04masked Hum
N N G N N N N N N
Cindy is pruning a plant. First, she cuts the longest branch to half of its original length. Then, she trims the leftmost branch until it is the same length as the branch next to it. Finally, she cuts off one-third of the rightmost branch's length.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
claude-fable-5-1 × × ✓ NARC —
gemma-4-26b × × ✓ Strong (0/3) Narrative (0/3)
gemma-4-31b ✓ × ✓ — —
gpt-oss-120b × × ✓ Strong (0/2) Partial (2/3)
gpt-oss-20b × × ✓ Strong (0/3) Narrative (0/1)
nemotron-3-super × × ✓ Strong (0/3) Narrative (0/3)
qwen3.5-122b × × ✓ Strong (0/3) Narrative (0/3)
qwen3.6-27b × × ✓ Strong (0/3) Narrative (0/3)
qwen3.8-27b × × ✓ Strong (0/3) Narrative (0/3)
▶ sub_065 🧩The Fountain Show 05 （3x4）01 masked Hum
N N × N × × × × ×
Different buttons control different water jet heights. By pressing different combinations of buttons, different fountain patterns can be created. During the rehearsal, Walter sets the water jet controlled by the first button to 1 meter, the second button to 2 meters, and the third button to 3 meters. During the first actual performance, he presses the second, first, and third buttons in sequence. During the second actual performance, he presses the third, first, and second buttons in sequence.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
claude-fable-5-1 × × ✓ NARC —
gemma-4-26b × × ✓ Strong (0/3) Narrative (0/3)
gemma-4-31b × × × — —
gpt-oss-120b × × ✓ Strong (0/3) Partial (1/3)
gpt-oss-20b × × × — —
nemotron-3-super × × × — —
qwen3.5-122b × × × — —
qwen3.6-27b × × × — —
qwen3.8-27b × × × — —
▶ sub_086 🧩Penny's car 01 small ver. Hum
N × × N N × N N N
Penny's car has a 100-liter fuel tank. She starts her trip with the tank full of 89-octane gasoline. After using half of the fuel, she refills the tank with 91-octane gasoline. Later, after using 40 liters of the mixed fuel, she refills the tank to full again with 91-octane gasoline. Penny wants to know how much 89-octane and 91-octane gasoline are in the tank now.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
claude-fable-5-1 × × ✓ NARC —
gemma-4-26b × × × — —
gemma-4-31b × × × — —
gpt-oss-120b × × ✓ Partial (2/3) Narrative (0/3)
gpt-oss-20b × × ✓ Strong (0/3) Narrative (0/3)
nemotron-3-super × × × — —
qwen3.5-122b × × ✓ Strong (0/3) Narrative (0/3)
qwen3.6-27b × × ✓ Strong (0/3) Narrative (0/2)
qwen3.8-27b × × ✓ Strong (0/3) Narrative (0/3)
▶ sub_088 🧩Moka pot 01 small ver. Hum
N × N N × × × N N
Bella was making coffee with a moka pot. She poured purified water into the bottom chamber and added ground coffee to the filter basket in the middle. As the pot heated up, pressure forced the water from the bottom chamber through the coffee grounds and into the top chamber, producing coffee. Not all of the water was transferred to the top chamber, a layer of water remained in the bottom chamber. Bella poured out all the coffee and began to enjoy it.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
claude-fable-5-1 × × ✓ NARC —
gemma-4-26b × × × — —
gemma-4-31b × × ✓ Strong (0/3) Narrative (0/3)
gpt-oss-120b × × ✓ Partial (2/3) Narrative (0/2)
gpt-oss-20b × × × — —
nemotron-3-super × × × — —
qwen3.5-122b × × × — —
qwen3.6-27b × × ✓ Strong (0/3) Lexical (3/3)
qwen3.8-27b × × ✓ Strong (0/3) Narrative (0/3)
▶ sub_090 🧩Carrot Patch 02 small ver. Hum
L × N N × × L N N
Garrett has a carrot patch. He pulls up one carrot at a time, skipping one carrot between each one he pulls. Unlike the previous two times, this time he left the carrot tops where they were.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
claude-fable-5-1 × ✓ × — —
gemma-4-26b × × × — —
gemma-4-31b × × ✓ Strong (0/3) Narrative (0/3)
gpt-oss-120b × × ✓ Strong (0/3) Narrative (0/3)
gpt-oss-20b × × × — —
nemotron-3-super × × × — —
qwen3.5-122b × ✓ ✓ — —
qwen3.6-27b × × ✓ Strong (0/3) Narrative (0/3)
qwen3.8-27b × × ✓ Strong (0/3) Narrative (0/3)
▶ sub_095 🧩Bubble Level 04（9x1） Hum
N × N N N × N N N
Mike is hanging a picture on the wall. He places a bubble level on top of the picture frame to make sure it is level. At first, the left side of the frame is higher, so the bubble in the bubble level shifts to the left. He then adjusts the frame, but overcorrects, causing the bubble to shift to the right. He makes a slight adjustment, but it is not enough, the right side is still higher. He adjusts the frame once more, and finally gets the picture perfectly level.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
claude-fable-5-1 × × ✓ NARC —
gemma-4-26b × × × — —
gemma-4-31b × × ✓ Partial (2/3) Partial (1/3)
gpt-oss-120b × × ✓ Weak (3/3) Partial (2/3)
gpt-oss-20b × × ✓ Partial (1/3) Narrative (0/2)
nemotron-3-super × × × — —
qwen3.5-122b × × ✓ Partial (2/3) Narrative (0/3)
qwen3.6-27b × × ✓ Partial (2/3) Narrative (0/3)
qwen3.8-27b × × ✓ Partial (1/3) Partial (1/3)
▶ sub_101 🧩Push-up 01 03,04masked Hum
N × × N × × G N N
Pamela is demonstrating how to do a push-up. She breaks down the movement into two steps and demonstrates it twice.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
claude-fable-5-1 × × ✓ NARC —
gemma-4-26b × × × — —
gemma-4-31b × × × — —
gpt-oss-120b × × ✓ Partial (1/3) Partial (1/3)
gpt-oss-20b × × × — —
nemotron-3-super × × × — —
qwen3.5-122b ✓ × ✓ — —
qwen3.6-27b × × ✓ Strong (0/3) Lexical (3/3)
qwen3.8-27b × × ✓ Strong (0/3) Partial (2/3)
▶ sub_103 🧩Six cakes01 03 masked Hum
× × L N × × × × ×
From above, Hannah sees six cakes on a table. After eating the two cakes on the right, she lowers the camera to cake level, where cakes behind others are hidden from view. She first takes a photo from the front and then another from the left side.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
claude-fable-5-1 × × × — —
gemma-4-26b × × × — —
gemma-4-31b × ✓ ✓ — —
gpt-oss-120b × × ✓ Strong (0/3) Narrative (0/3)
gpt-oss-20b × × × — —
nemotron-3-super × × × — —
qwen3.5-122b × × × — —
qwen3.6-27b × × × — —
qwen3.8-27b × × × — —
▶ sub_113 🧩Oil Hum
G × × N × × N G G
The ocean was once clean and filled with fish. But then an oil spill slowly seeped into the ocean. As the oil seeped down, it suffocated the fish it touched.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
claude-fable-5-1 ✓ × ✓ — —
gemma-4-26b × × × — —
gemma-4-31b × × × — —
gpt-oss-120b ✓ × ✓ — —
gpt-oss-20b × × × — —
nemotron-3-super × × × — —
qwen3.5-122b ✓ × × — —
qwen3.6-27b ✓ × ✓ — —
qwen3.8-27b ✓ × × — —
▶ sub_113_rev Oil2 Hum
G × × N × × × G N
The ocean was once clean and filled with fish. But then an oil spill slowly seeped into the ocean. As the oil seeped down, it suffocated the fish it touched.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
claude-fable-5-1 ✓ × ✓ — —
gemma-4-26b × × × — —
gemma-4-31b × × × — —
gpt-oss-120b ✓ × ✓ — —
gpt-oss-20b × × × — —
nemotron-3-super × × × — —
qwen3.5-122b × × × — —
qwen3.6-27b ✓ × × — —
qwen3.8-27b × × ✓ Strong (0/3) Partial (1/3)
▶ sub_114 Oil3 Hum
G × × N × × N × N
The ocean was once clean and filled with fish. But then an oil spill slowly seeped into the ocean. As the oil seeped down, it suffocated the fish it touched. The threat of the oil scared other fish into leaving as they saw their friends dead becuase of the oil.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
claude-fable-5-1 ✓ × ✓ — —
gemma-4-26b × × × — —
gemma-4-31b × × × — —
gpt-oss-120b × × ✓ Partial (1/3) Lexical (3/3)
gpt-oss-20b × × × — —
nemotron-3-super × × × — —
qwen3.5-122b × × × — —
qwen3.6-27b × × × — —
qwen3.8-27b × × ✓ Strong (0/3) Narrative (0/3)
▶ sub_117 SimpleTime Hum
N × N × × × × × ×
Counting like a digital clock.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
claude-fable-5-1 × × ✓ NARC —
gemma-4-26b × × × — —
gemma-4-31b × × × — —
gpt-oss-120b × × × — —
gpt-oss-20b × × × — —
nemotron-3-super × × × — —
qwen3.5-122b × × × — —
qwen3.6-27b × × × — —
qwen3.8-27b × × × — —
▶ sub_117_rev SimpleTimeBin Hum
G × G G G G N G N
Counting like a digital clock.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
claude-fable-5-1 ✓ × ✓ — —
gemma-4-26b × × × — —
gemma-4-31b ✓ × ✓ — —
gpt-oss-120b ✓ × ✓ — —
gpt-oss-20b ✓ × ✓ — —
nemotron-3-super ✓ × ✓ — —
qwen3.5-122b ✓ × ✓ — —
qwen3.6-27b ✓ × ✓ — —
qwen3.8-27b ✓ × ✓ — —
▶ synth_blueprint The Blueprint Colab
N L N N L N N N
The architect's plan called for offices filling the floor plate from the left, one bay at a time. By the fourth revision every bay was designated office space. Then the client changed the brief: the rightmost bay, the one added last and with the fewest structural commitments, was to become a server room instead. Same footprint, different function. The architect recolored that column on the plan from blue to red. The rest stayed blue. The final revision restored all bays to the same designation.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b × × ✓ Partial (1/3) Narrative (0/3)
gemma-4-31b × ✓ ✓ — —
gpt-oss-120b × × ✓ Weak (3/3) Partial (2/3)
gpt-oss-20b × × ✓ Strong (0/3) Narrative (0/3)
nemotron-3-super × ✓ ✓ — —
qwen3.5-122b × × ✓ Strong (0/3) Partial (1/3)
qwen3.6-27b × × ✓ Partial (1/3) Narrative (0/3)
qwen3.8-27b × × ✓ Partial (1/3) Partial (1/3)
▶ synth_chorus The Chorus Colab
N N N N × N × ×
They joined the chorus one row at a time from the front, the basses anchoring the sound. By the third rehearsal the whole ensemble was singing. Then the sopranos in the back row — the last section recruited, still sight-reading — lost their nerve and went silent. The rest held the chord.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b × × ✓ Partial (2/3) Narrative (0/3)
gemma-4-31b × × ✓ Partial (1/3) Narrative (0/3)
gpt-oss-120b × × ✓ Partial (1/3) Narrative (0/3)
gpt-oss-20b × × ✓ Strong (0/3) Narrative (0/3)
nemotron-3-super × × × — —
qwen3.5-122b × × ✓ Partial (1/3) Partial (2/3)
qwen3.6-27b × × × — —
qwen3.8-27b × × × — —
▶ synth_palimpsest The Palimpsest Colab
G G N N N G G G
The scribe copied the new text over the old, column by column from the left. But he was lazy about the last column — the parchment was thickest there, and the old ink showed through. Every other column was cleanly overwritten; the rightmost still bore the ghost of what had been there first.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
gemma-4-26b ✓ × ✓ — —
gemma-4-31b ✓ × ✓ — —
gpt-oss-120b × × ✓ Strong (0/3) Narrative (0/3)
gpt-oss-20b × × ✓ Partial (1/3) Narrative (0/3)
nemotron-3-super × × ✓ Partial (1/3) Partial (1/3)
qwen3.5-122b ✓ × ✓ — —
qwen3.6-27b ✓ × ✓ — —
qwen3.8-27b ✓ × × — —
▶ the Race_rev The Race Hum
G × × × × × × × ×
Juliet had been training for the Boston marathon for two years. Her goal: complete the race in under four hours. On the morning of the race, she lined up with all the other runners at the starting line. But when the starting buzzer sounded, an alien appeared in the sky. Strangely, Juliet was the only who saw the alien. While all the other runners started the race, Juliet remained glued to the spot, staring up at the sky. "Guys! Look! There's an alien!" she shouted. But no one listened. They all just kept running. Then the alien came closer, and closer, until it was hovering right in front of Juliet. "You have been chosen," said the alien. Then Juliet woke up.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
claude-fable-5-1 ✓ × ✓ — —
gemma-4-26b × × × — —
gemma-4-31b × × × — —
gpt-oss-120b × × × — —
gpt-oss-20b × × × — —
nemotron-3-super × × × — —
qwen3.5-122b × × × — —
qwen3.6-27b × × × — —
qwen3.8-27b × × × — —
▶ the date The Date Hum
N × × × × × × × ×
It was Deb's first date since her divorce. She met the man online, as so many people do these days, and after a couple weeks of correspondence, they decided to meet for dinner. He was perfectly pleasant. Polite, funny, relaxed. Nothing like her ex-husband, Chris. But no matter how hard she tried to enjoy the moment, to enjoy the date, the memories overtook her. She thought of her first date with Chris, so many years ago, right in this very restaurant. The memory was so strong, the nostalgia so consuming, that when she looked at the man across the table, all she saw was Chris. The new man sensed she wasn't ready to move on, and after he left, Deb sat at the table for a long time, staring at the empty chair.
Original narrative × original mask
Model Grids Only Narrative Only Both NARC Keywords
claude-fable-5-1 × × ✓ NARC —
gemma-4-26b × × × — —
gemma-4-31b × × × — —
gpt-oss-120b × × × — —
gpt-oss-20b × × × — —
nemotron-3-super × × × — —
qwen3.5-122b × × × — —
qwen3.6-27b × × × — —
qwen3.8-27b × × × — —