Open-set assistance on Overcooked Task Generalization (held-out)
85.96Coaching Score (LLM-as-Judge)Open-Set Corrective Assistance
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| Open-Set Corrective AssistanceModel Scale=8B2026.03 | 85.96 | 56.67 | |
| Behavior Critic + SummariesSummaries usage=true2026.03 | 71.05 | 15.83 | |
| Open-Set Corrective AssistanceModel Scale=1B2026.03 | 50.88 | 50.83 | |
| Behavior Critic2026.03 | 34.21 | 9.17 |