Health Multimodal Evaluation on HealthBench English (test)
32.7Overall ScorePOW3R dynamic
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| POW3R dynamicBase policy=Qwen3-4B, Setting=POW3R dynamic2026.05 | 32.7 | 4.7 | 5.4 | |
| POW3R dynamicBase policy=Qwen3-8B, Setting=POW3R dynamic2026.05 | 31.5 | 3.7 | 4.6 | |
| Category-balancedBase policy=Qwen3-8B, Setting=Category-balanced2026.05 | 30.7 | 2.9 | 4.8 | |
| Category-balancedBase policy=Qwen3-4B, Setting=Category-balanced2026.05 | 30.5 | 2.5 | 4.6 | |
| Static scalarBase policy=Qwen3-8B, Setting=Static scalar2026.05 | 29.9 | 2.1 | 4.5 | |
| Static scalarBase policy=Qwen3-4B, Setting=Static scalar2026.05 | 29.6 | 1.6 | 3.8 | |
| POW3R dynamicBase policy=Gemma3-4B, Setting=POW3R dynamic2026.05 | 28.1 | 3.8 | 2.7 | |
| BaseBase policy=Qwen3-4B, Setting=Base2026.05 | 28 | — | 4.2 | |
| BinaryBase policy=Qwen3-8B, Setting=Binary2026.05 | 27.9 | 0.1 | 4.4 | |
| BinaryBase policy=Qwen3-4B, Setting=Binary2026.05 | 27.8 | 0.2 | 4 | |
| BaseBase policy=Qwen3-8B, Setting=Base2026.05 | 27.8 | — | 4.4 | |
| Category-balancedBase policy=Gemma3-4B, Setting=Category-balanced2026.05 | 27.1 | 2.8 | 2.6 | |
| Static scalarBase policy=Gemma3-4B, Setting=Static scalar2026.05 | 25.4 | 1.1 | 2.8 | |
| BaseBase policy=Gemma3-4B, Setting=Base2026.05 | 24.3 | — | 2.4 | |
| BinaryBase policy=Gemma3-4B, Setting=Binary2026.05 | 24.2 | 0.1 | 2.2 |