STARFlow2: Bridging Language Models and Normalizing Flows for Unified Multimodal Generation

Open paper
Benchmark Model Rank Results
multimodal-reasoning-on-mmmuSTARFlow2 (10.6B)–Accuracy: 67.7
text-to-image-generation-on-dpg-benchSTARFlow2 (10.6B)–DPG-Bench Score: 84.94
text-to-image-generation-on-genevalSTARFlow2 (10.6B)–Overall: 0.82

Rank counts only results with a code link.