Segment-level mistake detection on CaptainCook4D 5-recipe average
16.2F1@.5GTG2Vid
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| GTG2VidEvaluation Protocol=Fully supervised2026.06 | 16.2 | 66.3 | |
| ZeProM (Q-L)Evaluation Protocol=Zero-shot, Backbone=Qwen3.5-397B-A17B-GPTQ-Int42026.06 | 16 | 67.4 | |
| EgoPEDEvaluation Protocol=Fully supervised2026.06 | 10.5 | — | |
| ZeProM (Q-M)Evaluation Protocol=Zero-shot, Backbone=Qwen3.5-35B-A3B2026.06 | 10.2 | 66.3 |