S1 flipped a pancake from one video prompt; not in pre-training data
The first time S1 flipped a pancake, the team assumed pancake flipping must have been in its pre-training data.
They searched the whole pre-training data and found no examples of flipping.
S1 inferred the out-of-distribution task from one video prompt.
▶
✓ Verified
2026-08-25