Multimodal Mathematical Reasoning on MMMU
54.11AccuracyCEPO
Evaluation Results
| Method | Links | |
|---|---|---|
| CEPOBackbone=Qwen3-VL-4B-Instruct2026.05 | 54.11 | |
| BaseBackbone=Qwen3-VL-4B-Instruct2026.05 | 53.56 | |
| RLSDBackbone=Qwen3-VL-4B-Instruct2026.05 | 53.22 | |
| GRPOBackbone=Qwen3-VL-4B-Instruct2026.05 | 52.34 | |
| SDPOBackbone=Qwen3-VL-4B-Instruct2026.05 | 48.11 | |
| OPSDBackbone=Qwen3-VL-4B-Instruct2026.05 | 47 | |
| CEPOBackbone=Qwen3-VL-2B-Instruct2026.05 | 45.78 | |
| BaseBackbone=Qwen3-VL-2B-Instruct2026.05 | 44.11 | |
| OPSDBackbone=Qwen3-VL-2B-Instruct2026.05 | 43.78 | |
| SDPOBackbone=Qwen3-VL-2B-Instruct2026.05 | 43 | |
| GRPOBackbone=Qwen3-VL-2B-Instruct2026.05 | 42.33 | |
| RLSDBackbone=Qwen3-VL-2B-Instruct2026.05 | 39.44 |