Multimodal Instruction Following Evaluation on MM-IFE
14.8MM-IFE ScorePOW3R dynamic
Evaluation Results
| Method | Links | |
|---|---|---|
| POW3R dynamicBase policy=Qwen3-VL-8B2026.05 | 14.8 | |
| Category-balancedBase policy=Qwen3-VL-8B2026.05 | 14.2 | |
| POW3R dynamicBase policy=Qwen3-VL-4B2026.05 | 13.5 | |
| Static scalarBase policy=Qwen3-VL-8B2026.05 | 13.5 | |
| Category-balancedBase policy=Qwen3-VL-4B2026.05 | 13 | |
| Static scalarBase policy=Qwen3-VL-4B2026.05 | 12.5 | |
| BaseBase policy=Qwen3-VL-8B2026.05 | 12.5 | |
| BinaryBase policy=Qwen3-VL-8B2026.05 | 12.3 | |
| BinaryBase policy=Qwen3-VL-4B2026.05 | 11.8 | |
| BaseBase policy=Qwen3-VL-4B2026.05 | 11.5 | |
| POW3R dynamicBase policy=Gemma3-4B2026.05 | 7.2 | |
| Category-balancedBase policy=Gemma3-4B2026.05 | 6.7 | |
| Static scalarBase policy=Gemma3-4B2026.05 | 6.3 | |
| BinaryBase policy=Gemma3-4B2026.05 | 6.1 | |
| BaseBase policy=Gemma3-4B2026.05 | 6 |