Closed-world plan-synthesis on ALFWorld unseen (val)
100Success Rate (SR)Fast Downward (upper bound)
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| Fast Downward (upper bound)2026.07 | 100 | — | |
| GPT-5.5 (LLM+VAL loop)Backbone=GPT-5.52026.07 | 100 | 2,770,945 | |
| EvoPlanBackbone=Qwen3-32B2026.07 | 90.3 | 10,349,634 | |
| EvoPlanBackbone=gpt-oss-120b2026.07 | 78.4 | 17,071,316 |