General AI Assistant on GAIA text
70.5GAIA Average ScoreOpenAI-o3
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| OpenAI-o32026.01 | 70.5 | — | — | — | |
| Claude-4-Sonnet2026.01 | 68.3 | — | — | — | |
| OpenAI DeepResearch2026.01 | 67.4 | — | — | — | |
| ARPOBackbone=Qwen3-14B2026.01 | 43.7 | 56.4 | 40.4 | 16.7 | |
| Reagent-UBackbone=Qwen3-8B2026.01 | 43.7 | 59 | 38.5 | 16.7 | |
| WebDancerBackbone=Qwen2.5-32B2026.01 | 40.7 | 46.1 | 44.2 | 8.3 | |
| Search-o1Backbone=QwQ-32B-Preview2026.01 | 39.8 | 53.8 | 34.6 | 16.7 | |
| ARPOBackbone=Qwen3-8B2026.01 | 38.8 | 53.9 | 32.7 | 16.7 | |
| Reagent-RBackbone=Qwen3-8B2026.01 | 36.9 | 51.3 | 30.8 | 16.7 | |
| VerlToolBackbone=Qwen3-8B2026.01 | 34 | — | — | — | |
| Reagent w/o Agent-RRMBackbone=Qwen3-8B2026.01 | 34 | 41 | 36.5 | 0 | |
| WebDancerBackbone=Qwen2.5-7B2026.01 | 31 | 41 | 30.7 | 0 | |
| DeepSeek-R1-671BBackbone=DeepSeek-R1-671B2026.01 | 25.2 | 40.5 | 21.2 | 5.2 | |
| Reagent-CBackbone=Qwen3-8B, Inference Type=Direct Inference2026.01 | 25.2 | 30.8 | 23.1 | 16.7 | |
| WebThinkerBackbone=Qwen3-8B2026.01 | 22.3 | 43.6 | 11.5 | 0 | |
| Qwen3-8BBackbone=Qwen3-8B2026.01 | 21.4 | 28.2 | 21.2 | 0 | |
| Tree-GRPOBackbone=Qwen2.5-14B2026.01 | 21 | 20.8 | 24.3 | 7.3 | |
| QwQ-32BBackbone=QwQ-32B2026.01 | 18.9 | 30.9 | 6.5 | 5.2 | |
| Atom-SearcherBackbone=Qwen2.5-7B2026.01 | 17.5 | 18 | 21.2 | 0 |