Interactive Decision-making on ScienceWorld Unseen (test)
58.94Success RateITPR
Evaluation results
| Method | Links | |
|---|---|---|
| ITPRBackbone=Qwen2.5-7B, Type=Training2026.01 | 58.94 | |
| IWMBackbone=Qwen2.5-7B, Type=Training2026.01 | 57.61 | |
| ITPRBackbone=Llama3.1-8B, Type=Training2026.01 | 57.61 | |
| ITPRBackbone=Qwen3-8B, Type=Training2026.01 | 56.95 | |
| WKMBackbone=Qwen2.5-7B, Type=Training2026.01 | 56.29 | |
| IWMBackbone=Llama3.1-8B, Type=Training2026.01 | 56.29 | |
| WKMBackbone=Llama3.1-8B, Type=Training2026.01 | 54.96 | |
| IWMBackbone=Qwen3-8B, Type=Training2026.01 | 54.3 | |
| SFTBackbone=Llama3.1-8B, Type=Training2026.01 | 50.33 | |
| SFTBackbone=Qwen3-8B, Type=Training2026.01 | 49.67 | |
| SFTBackbone=Qwen2.5-7B, Type=Training2026.01 | 49 | |
| WKMBackbone=Qwen3-8B, Type=Training2026.01 | 47.68 | |
| RAPBackbone=Qwen3-8B, Type=Prompting2026.01 | 27.14 | |
| ITPIBackbone=Qwen3-8B, Type=Prompting2026.01 | 19.86 | |
| ITPIBackbone=Llama3.1-8B, Type=Prompting2026.01 | 19.2 | |
| ITPIBackbone=Qwen2.5-7B, Type=Prompting2026.01 | 17.88 | |
| RAPBackbone=Llama3.1-8B, Type=Prompting2026.01 | 17.21 | |
| RAPBackbone=Qwen2.5-7B, Type=Prompting2026.01 | 16.55 | |
| ReActBackbone=Llama3.1-8B, Type=Prompting2026.01 | 13.24 | |
| ReActBackbone=Qwen2.5-7B, Type=Prompting2026.01 | 9.93 | |
| ReActBackbone=Qwen3-8B, Type=Prompting2026.01 | 8.61 | |
| CoTBackbone=Qwen2.5-7B, Type=Prompting2026.01 | 4.63 | |
| CoTBackbone=Llama3.1-8B, Type=Prompting2026.01 | 3.31 | |
| CoTBackbone=Qwen3-8B, Type=Prompting2026.01 | 1.99 |