Sparse-LaViDa: Sparse Multimodal Discrete Diffusion Language Models

Benchmark Model Rank Results
chart-question-answering-on-chartqaSparse-LaViDa–1:1 Accuracy: 82.0
image-editing-on-imgedit-dataSparse-LaViDa–Overall: 3.79Add: 4.08Adjust: 3.73Extract: 2.10Replace: 4.29…
mathematical-reasoning-on-mathverseSparse-LaViDa–Accuracy: 37.9
multimodal-reasoning-on-mathvistaSparse-LaViDa–Accuracy: 56.7
multimodal-reasoning-on-mmmuSparse-LaViDa–Accuracy: 43.6
text-to-image-generation-on-genevalSparse-LaViDa–Overall: 0.78