Question Answering on DoRA Gold Evidence Context
70.6Token F1DoRA SFT
Evaluation results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| DoRA SFT2026.04 | 70.6 | 71.5 | 57.3 | 0.224 | 93.2 | |
| Mistral-7B-Instruct2026.04 | 59.2 | 58.4 | 38 | -0.004 | 90.32 | |
| Llama3-8B-Instruct2026.04 | 47.1 | 49 | 18 | -0.231 | 86.88 | |
| Llama3.1-8B-Instruct2026.04 | 46.3 | 48.1 | 17 | -0.227 | 86.63 | |
| GPT-4o2026.04 | 45.6 | 46.7 | 14.2 | -0.088 | 86.92 | |
| GPT-3.5-Turbo2026.04 | 43.8 | 44.7 | 15.3 | -0.099 | 87.09 | |
| GPT-4o-mini2026.04 | 35.5 | 36.4 | 9.2 | -0.196 | 83.85 | |
| Qwen2.5-7B-Instruct2026.04 | 30.6 | 32 | 7.9 | -0.311 | 82.25 | |
| Claude-3-Haiku2026.04 | 29.3 | 30.3 | 10.4 | -0.381 | 82.99 |