Embodied Planning on AlfWorld (out of distribution)
58.95AccuracyProCeedSFT
Evaluation Results
| Method | Links | |
|---|---|---|
| ProCeedSFTBackbone=Qwen3-8B2026.04 | 58.95 | |
| ProCeedRLBackbone=Qwen3-8B2026.04 | 55.22 | |
| DAPOBackbone=Qwen3-8B2026.04 | 53.24 | |
| RFTBackbone=Qwen3-8B2026.04 | 50.25 | |
| Qwen3-8BBackbone=Qwen3-8B2026.04 | 47.07 | |
| DAPOBackbone=Qwen3-1.7B2026.04 | 24.12 | |
| ProCeedRLBackbone=Qwen3-1.7B2026.04 | 23.33 | |
| Qwen3-1.7BBackbone=Qwen3-1.7B2026.04 | 12.69 |