Multi-discipline Visual Reasoning on MMMU (val)
29.44AccuracyMMMaDA + VISAGE
Evaluation results
| Method | Links | |
|---|---|---|
| MMMaDA + VISAGEBase Model=MMMaDA, Decoding Strategy=VISAGE2026.03 | 29.44 | |
| MMMaDA + VCDBase Model=MMMaDA, Decoding Strategy=VCD2026.03 | 28.44 | |
| MMMaDA (Base)Base Model=MMMaDA2026.03 | 27.11 |