Question Answering on GAIA (held-out)
47.7EMWebExpert+SFT
Evaluation Results
| Method | Links | |
|---|---|---|
| WebExpert+SFTBackbone=QwQ-32B, Training=SFT2026.02 | 47.7 | |
| WebExpertBackbone=QwQ-32B2026.02 | 46.2 | |
| WebThinker-32B-BaseBackbone=32B-Base2026.02 | 44.7 | |
| Search-o1-32BBackbone=o1-32B2026.02 | 39.8 | |
| RAG-QwQ-32BStrategy=RAG2026.02 | 32 | |
| QwQ-32BStrategy=Direct2026.02 | 13.6 |