Multi-hop Question Answering on Musique (dev test)
49.2F1HiExp-Searcher
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| HiExp-SearcherEvaluation Protocol=Training Based, Base Model=Qwen2.5-32B2026.04 | 49.2 | 41.1 | 36.2 | |
| GPT-4.1Evaluation Protocol=Prompt Based, Category=Frontier LLMs2026.04 | 44.9 | 47 | 28.5 | |
| Qwen3-235B-A22BEvaluation Protocol=Prompt Based, Category=Frontier LLMs2026.04 | 41.7 | 39.5 | 27.6 | |
| HiExp-SearcherEvaluation Protocol=Training Based, Base Model=Qwen2.5-7B2026.04 | 41.7 | 36.7 | 30.7 | |
| o4-miniEvaluation Protocol=Prompt Based, Category=Frontier LLMs2026.04 | 41.6 | 45.5 | 27.5 | |
| DeepSeek-R1Evaluation Protocol=Prompt Based, Category=Frontier LLMs2026.04 | 39.9 | 33 | 27.5 | |
| ReSearchEvaluation Protocol=Training Based, Base Model=Qwen2.5-32B2026.04 | 39.3 | 33.8 | 30.5 | |
| Gemini-2.5-ProEvaluation Protocol=Prompt Based, Category=Frontier LLMs2026.04 | 37 | 47 | 24.5 | |
| Search-R1-v0.3Evaluation Protocol=Training Based, Base Model=Qwen2.5-32B2026.04 | 36.2 | 30.6 | 28.5 | |
| Search-R1-v0.3Evaluation Protocol=Training Based, Base Model=Qwen2.5-7B2026.04 | 30.9 | 24.7 | 21.5 | |
| R1-SearcherEvaluation Protocol=Training Based, Base Model=Qwen2.5-7B2026.04 | 28.4 | 27.9 | 19.5 | |
| ReSearchEvaluation Protocol=Training Based, Base Model=Qwen2.5-7B2026.04 | 28 | 34.1 | 24 | |
| Search-o1 + HiExpEvaluation Protocol=Prompt Based, Base Model=Qwen2.5-7B2026.04 | 22.4 | 18.6 | 15.1 | |
| Search-o1Evaluation Protocol=Prompt Based, Base Model=Qwen2.5-7B, Status=Reproduced2026.04 | 18.1 | 15.5 | 11.1 | |
| Iter-RetGenEvaluation Protocol=Prompt Based, Base Model=Qwen2.5-7B2026.04 | 17.4 | 12.4 | 10 | |
| IRCoTEvaluation Protocol=Prompt Based, Base Model=Qwen2.5-7B2026.04 | 14.7 | 13.3 | 7.5 | |
| Vanilla RAGEvaluation Protocol=Prompt Based, Base Model=Qwen2.5-7B2026.04 | 11.2 | 5.1 | 3.4 |