Clinical Task Execution on MedAgentBench OOD
80.6AccuracyGRASP
Evaluation Results
| Method | Links | |
|---|---|---|
| GRASPModel=GPT-5.4 (low)2026.05 | 80.6 | |
| GRASPModel=gpt-oss-120b2026.05 | 56.3 | |
| GRASPModel=DeepSeek V4 Flash2026.05 | 52.3 | |
| GRASPModel=Gemini 3.1 Flash Lite2026.05 | 41.7 | |
| SkillXModel=Gemini 3.1 Flash Lite2026.05 | 38.3 | |
| Evo-MedAgentModel=gpt-oss-120b2026.05 | 31.3 | |
| GRASPModel=GPT-4.12026.05 | 27.2 | |
| SkillXModel=gpt-oss-120b2026.05 | 21.3 | |
| Evo-MedAgentModel=DeepSeek V4 Flash2026.05 | 20 | |
| No skillsModel=Gemini 3.1 Flash Lite2026.05 | 18.1 | |
| ExpeLModel=Gemini 3.1 Flash Lite2026.05 | 17.2 | |
| Seq. MemoryModel=gpt-oss-120b2026.05 | 16.7 | |
| Batch MemoryModel=GPT-5.4 (low)2026.05 | 16.7 | |
| Evo-MedAgentModel=Gemini 3.1 Flash Lite2026.05 | 15.6 | |
| Seq. MemoryModel=GPT-5.4 (low)2026.05 | 15.6 | |
| Batch MemoryModel=gpt-oss-120b2026.05 | 13 | |
| ExpeLModel=gpt-oss-120b2026.05 | 12 | |
| ExpeLModel=GPT-5.4 (low)2026.05 | 11.7 | |
| Seq. MemoryModel=Gemini 3.1 Flash Lite2026.05 | 8.9 | |
| No skillsModel=gpt-oss-120b2026.05 | 8.7 | |
| Seq. MemoryModel=DeepSeek V4 Flash2026.05 | 8.7 | |
| Batch MemoryModel=Gemini 3.1 Flash Lite2026.05 | 8.3 | |
| SkillXModel=DeepSeek V4 Flash2026.05 | 6.3 | |
| Batch MemoryModel=DeepSeek V4 Flash2026.05 | 3.3 | |
| SkillXModel=GPT-5.4 (low)2026.05 | 3.3 | |
| No skillsModel=DeepSeek V4 Flash2026.05 | 3.1 | |
| ExpeLModel=DeepSeek V4 Flash2026.05 | 2.3 | |
| Batch MemoryModel=GPT-4.12026.05 | 2.2 | |
| No skillsModel=GPT-5.4 (low)2026.05 | 1.7 | |
| Evo-MedAgentModel=GPT-5.4 (low)2026.05 | 1.7 | |
| Seq. MemoryModel=GPT-4.12026.05 | 0.6 | |
| ExpeLModel=GPT-4.12026.05 | 0.6 | |
| Evo-MedAgentModel=GPT-4.12026.05 | 0.6 | |
| No skillsModel=GPT-4.12026.05 | 0.3 | |
| SkillXModel=GPT-4.12026.05 | 0 |