Question Answering on Neuroscience QA 5-hop
88AccuracyQwen 14B
Evaluation Results
| Method | Links | |
|---|---|---|
| Qwen 14BTraining Phase=SFT + RL2026.05 | 88 | |
| Qwen 14BTraining Phase=SFT2026.05 | 84.8 | |
| Gemini 3.1 ProModel=Gemini 3.1 Pro2026.05 | 82.2 | |
| Qwen 14BTraining Phase=Base2026.05 | 76.4 |