PixelUMM: Encoder-Free Unified Image and Video Understanding and Generation

Benchmark Model Rank Results
text-to-video-generation-on-vbenchPixelUMM 8B MoT#25VBench Total Score: 83.24VBench Quality Score: 84.1
video-question-answering-on-mvbenchPixelUMM 8B MoT#27Avg.: 70.53
visual-question-answering-vqa-on-blinkPixelUMM 8B MoT#41Accuracy: 53.46
zero-shot-video-question-answer-on-video-mme-w-o-subsPixelUMM 8B MoT#49Accuracy (%): 57.33
video-question-answering-on-lvbenchPixelUMM 8B MoT#50Accuracy: 40.41
chart-question-answering-on-chartqaPixelUMM 8B MoT#561:1 Accuracy: 82.96
multimodal-reasoning-on-mmmu-proPixelUMM 8B MoT#60Accuracy: 27.63
vision-language-understanding-on-mmstarPixelUMM 8B MoT#91Accuracy: 53.99
text-to-image-generation-on-genevalPixelUMM 8B MoT#118Overall: 0.77
multimodal-reasoning-on-mmmuPixelUMM 8B MoT#121Accuracy: 41.67