Agentic Search on MuSiQue
32.5String F1 ScoreVanilla-Qwen3-32B
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| Vanilla-Qwen3-32BASP=false2026.04 | 32.5 | — | — | — | |
| Vanilla-Qwen3-8BASP=false2026.04 | 32 | — | — | — | |
| SFT-Qwen3-1.7BASP=true2026.04 | 29.2 | — | — | — | |
| Vanilla-Qwen3-4BASP=false2026.04 | 28.7 | — | — | — | |
| OPD-Qwen3-1.7BASP=true2026.04 | 28.6 | — | — | — | |
| Mixed-Qwen3-1.7BASP=true2026.04 | 27.2 | — | — | — | |
| Distilled-Qwen3-1.7BASP=false2026.04 | 25.4 | — | — | — | |
| Vanilla-Qwen3-1.7BASP=false2026.04 | 22.9 | — | — | — | |
| SFT-Qwen3-0.6BASP=true2026.04 | 22.7 | — | — | — | |
| SFT-Llama-3.2-1BASP=true2026.04 | 22.3 | — | — | — | |
| Distilled-Llama3.2-3BASP=false2026.04 | 20.6 | — | — | — | |
| Distilled-Llama3.2-1BASP=false2026.04 | 15.9 | — | — | — | |
| SFT-Llama-3.2-3BASP=true2026.04 | 14.4 | — | — | — | |
| Vanilla-Qwen3-0.6BASP=false2026.04 | 8.7 | — | — | — | |
| LaserBase Model=Qwen3-8B2025.12 | — | 22.4 | 28.2 | 20.4 | |
| LaserBase Model=Qwen3-32B2025.12 | — | 24.2 | 31.6 | 22 | |
| No-thinkingBase Model=Qwen3-8B2025.12 | — | 4.2 | 6 | 3 | |
| No-thinkingBase Model=Qwen3-32B2025.12 | — | 7.8 | 11.2 | 5.8 | |
| RAGBase Model=Qwen3-8B2025.12 | — | 9 | 10.8 | 7.4 | |
| RAGBase Model=Qwen3-32B2025.12 | — | 12.2 | 14 | 10 | |
| ReActBase Model=Qwen3-8B2025.12 | — | 19.4 | 24.6 | 14.6 | |
| ReActBase Model=Qwen3-32B2025.12 | — | 22 | 26.8 | 15.2 | |
| Search-R1Base Model=Qwen3-8B2025.12 | — | 10.4 | 14.6 | 7.6 | |
| Search-R1Base Model=Qwen3-32B2025.12 | — | 17.6 | 22.2 | 11.4 | |
| ThinkingBase Model=Qwen3-8B2025.12 | — | 5.8 | 8.6 | 3.4 | |
| ThinkingBase Model=Qwen3-32B2025.12 | — | 10.6 | 14 | 7.8 |