Biological Reasoning on BioAlchemy
52.78ProtocolQA AccuracyGPT-OSS-20B
Evaluation Results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| GPT-OSS-20B2026.04 | 52.78 | 18.12 | 7.27 | 72.79 | 55.79 | 41.35 | |
| BioAlchemist-8BTraining Data Size=150K2026.04 | 46.2 | 22.32 | 15.15 | 68.32 | 62.11 | 42.82 | |
| BioAlchemist-8BTraining Data Size=50K2026.04 | 45.83 | 18.73 | 11.52 | 68.09 | 57.89 | 40.41 | |
| Qwen3-8B2026.04 | 42.69 | 8.42 | 5.76 | 69 | 42.63 | 33.7 | |
| DeepSeek-R1-Llama-8B2026.04 | 33.61 | 4.97 | 10.91 | 26.89 | 5.26 | 16.33 |