Text-based embodied task on Jericho
15Success RateQwen2.5-72B
Evaluation results
| Method | Links | ||
|---|---|---|---|
| Qwen2.5-72BActor=Qwen2.5-72B, Thinker=No2026.05 | 15 | 30.87 | |
| GPT-4oActor=GPT-4o, Thinker=No2026.05 | 10 | 34 | |
| AgentRefineActor=AgentRefine, Thinker=No2026.05 | 10 | 32.3 | |
| LLaMA3-8BActor=LLaMA3-8B, Thinker=Exp-Thinker2026.05 | 5 | 26.31 | |
| Qwen2.5-7BActor=Qwen2.5-7B, Thinker=Exp-Thinker2026.05 | 5 | 19.94 | |
| Agent-FLANActor=Agent-FLAN, Thinker=No2026.05 | 0 | 10.1 | |
| AgentGymActor=AgentGym, Thinker=No2026.05 | 0 | 12.9 | |
| LLaMA3-8BActor=LLaMA3-8B, Thinker=No2026.05 | 0 | 14.23 | |
| Qwen2.5-7BActor=Qwen2.5-7B, Thinker=No2026.05 | 0 | 8.95 | |
| LLaMA3-ActorActor=LLaMA3-Actor, Thinker=No2026.05 | 0 | 15.98 | |
| LLaMA3-ActorActor=LLaMA3-Actor, Thinker=Exp-Thinker2026.05 | 0 | 17.11 | |
| Qwen2.5-ActorActor=Qwen2.5-Actor, Thinker=No2026.05 | 0 | 17.44 | |
| Qwen2.5-ActorActor=Qwen2.5-Actor, Thinker=Exp-Thinker2026.05 | 0 | 17.64 |