Long-Horizon Stability on BioProBench Subset C (test)
100Success RateBioProAgent
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| BioProAgentBackbone=Gemini-3-Flash2026.03 | 100 | 71.8 | 95 | |
| ReActBackbone=Gemini-3-Flash2026.03 | 88.9 | 11.4 | 21.7 | |
| AutoGPTBackbone=Gemini-3-Flash2026.03 | 66.7 | 40.9 | 64.4 | |
| ReflexionBackbone=Gemini-3-Flash2026.03 | 33.3 | 0 | 0 |