General AI Assistant Task Execution on GAIA Round 2
20Level 1 Evo. Pass@1Deepseek-v3.1
Evaluation Results
| Method | Links | ||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Deepseek-v3.12025.10 | 20 | 50.9 | -0.309 | 22.2 | 41.9 | -0.197 | 0 | 23.1 | -0.231 | 18.8 | 41.8 | -0.229 | |
| GPT-5-Mini2025.10 | 16 | 56.6 | -0.406 | 13.9 | 46.5 | -0.326 | 37.5 | 19.2 | 0.183 | 17.4 | 45.5 | -0.281 | |
| KIMI-K22025.10 | 8 | 37.7 | -0.297 | 25 | 23.3 | 0.017 | 12.5 | 7.7 | 0.048 | 17.4 | 25.5 | -0.081 | |
| Gemini-2.5-flash2025.10 | 4 | 47.1 | -0.431 | 19.4 | 23.2 | -0.038 | 12.5 | 11.5 | 0.01 | 13 | 29.1 | -0.161 |