Agentic Search on BrowseComp (String-F1)
21String-F1Vanilla-Qwen3-32B
Evaluation results
| Method | Links | |
|---|---|---|
| Vanilla-Qwen3-32BASP=false2026.04 | 21 | |
| Vanilla-Qwen3-8BASP=false2026.04 | 16 | |
| Distilled-Qwen3-1.7BASP=false2026.04 | 10.1 | |
| SFT-Qwen3-1.7BASP=true2026.04 | 10 | |
| SFT-Llama-3.2-3BASP=true2026.04 | 10 | |
| SFT-Qwen3-0.6BASP=true2026.04 | 9 | |
| Mixed-Qwen3-1.7BASP=true2026.04 | 8.8 | |
| OPD-Qwen3-1.7BASP=true2026.04 | 8.4 | |
| SFT-Llama-3.2-1BASP=true2026.04 | 8.2 | |
| Vanilla-Qwen3-4BASP=false2026.04 | 7 | |
| Distilled-Llama3.2-1BASP=false2026.04 | 5.1 | |
| Vanilla-Qwen3-0.6BASP=false2026.04 | 4.6 | |
| Distilled-Llama3.2-3BASP=false2026.04 | 3.5 | |
| Vanilla-Qwen3-1.7BASP=false2026.04 | 3.2 |