Scientific Reasoning on Scientific Reasoning Subset A
14.7ROUGE-LBioProAgent
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| BioProAgentBackbone=Gemini-3-Flash2026.03 | 14.7 | 34.4 | 59.1 | 71.8 | |
| DirectBackbone=Gemini-3-Flash2026.03 | 13 | 24.7 | 32.2 | 12.1 | |
| DirectBackbone=DeepSeek-V32026.03 | 12.3 | 26 | 28.5 | 52.1 | |
| ReflexionBackbone=Gemini-3-Flash2026.03 | 11.8 | 28.2 | 43.9 | 148.4 | |
| ReActBackbone=Gemini-3-Flash2026.03 | 11.6 | 26.8 | 45.5 | 44.5 | |
| AutoGPTBackbone=Gemini-3-Flash2026.03 | 11.6 | 25.8 | 42.9 | 119.6 | |
| DirectBackbone=GPT-4o2026.03 | 10.7 | 20.2 | 18.9 | 13.8 | |
| BiomniBackbone=(Specialized)2026.03 | 8.1 | 25.2 | 34.2 | 87.1 |