MEND: RL for Flow Models via Proximal Velocity Matching Collection Official MEND paper, SD3.5 Medium LoRA weights, code and blog. PickScore at 100 updates; three-reward alignment at 300 updates. • 11 items • Updated 14 days ago • 1
Diffusion Research Collection Research work on Guidance and test-time search for diffusion models, efficient sampling, post-training, RL, perceptual quality,HDR-UGC video quality, • 22 items • Updated 8 days ago • 1