Text-based embodied skill selection on ALFWorld
93.6Success RateGoS
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| GoSBackbone Model=gpt-5.2-codex2026.06 | 93.6 | 86.4 | 56.4 | 67.9 | |
| SkillDAGBackbone Model=gpt-5.2-codex2026.06 | 93.6 | 85.1 | 60.4 | 69.6 | |
| Vector SkillsBackbone Model=gpt-5.2-codex2026.06 | 92.9 | 68.6 | 37.9 | 49.2 | |
| Vanilla SkillsBackbone Model=gpt-5.2-codex2026.06 | 89.3 | — | — | — | |
| SkillDAGBackbone Model=MiniMax-M2.72026.06 | 67.1 | 92.1 | 57.9 | 71.1 | |
| GoSBackbone Model=MiniMax-M2.72026.06 | 54.3 | 86.4 | 56.4 | 67.9 | |
| Vector SkillsBackbone Model=MiniMax-M2.72026.06 | 50.7 | 68.6 | 37.9 | 49.2 | |
| Vanilla SkillsBackbone Model=MiniMax-M2.72026.06 | 47.1 | — | — | — |