Data Research Agent Evaluation on DRACO 50 tasks sample
69.36Normalized ScoreDeepSeek-V4-Flash
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| DeepSeek-V4-FlashSetting=RGI Turn 32026.06 | 69.36 | 74.59 | 31.52 | 8.96 | |
| DeepSeek-V4-FlashSetting=RGI Turn 22026.06 | 65.35 | 71.1 | 39.61 | 13.41 | |
| DeepSeek-V4-FlashSetting=Turn 12026.06 | 57.2 | 63.94 | — | — | |
| DeepSeek-V4-FlashSetting=SR Turn 22026.06 | 56.66 | 63.84 | 26.18 | 15.99 | |
| GPT-4.1Setting=RGI Turn 22026.06 | 56.19 | 62.22 | 36.88 | 16.87 | |
| GPT-4.1-miniSetting=RGI Turn 32026.06 | 54.45 | 60.92 | 27.46 | 18.59 | |
| GPT-4.1-miniSetting=RGI Turn 22026.06 | 53.11 | 59.91 | 34.78 | 14.52 | |
| GPT-4.1Setting=RGI Turn 32026.06 | 51.22 | 58.86 | 27.17 | 23.57 | |
| GPT-4.1Setting=SR Turn 22026.06 | 44.86 | 51.94 | 15.58 | 14.74 | |
| GPT-4.1Setting=Turn 12026.06 | 44.77 | 51.55 | — | — | |
| GPT-4.1-miniSetting=SR Turn 22026.06 | 40.18 | 48.64 | 15.4 | 12.9 | |
| GPT-4.1-miniSetting=Turn 12026.06 | 37.76 | 45.89 | — | — |