You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
Paper suggestion: MEND: RL For Flow Models via Proximal Velocity Matching #12
I am an author of MEND: RL For Flow Models via Proximal Velocity Matching (arXiv preprint, October 2026). Please consider it for the manually curated Applications section.
MEND studies reward post-training of flow-matching image generators using proximal velocity matching. It constructs reward-improving velocity targets and fits them by regression. It requires a differentiable reward, and the reported experiments are text-to-image generation.
Related PR: #11. This issue tracks the same paper suggestion; the PR already contains the proposed changes, and only one listing is intended.
Proposed entry:
README.md:
**MEND: RL For Flow Models via Proximal Velocity Matching**\*Shreshth Saini, Neil Birkbeck, Yilin Wang, Balu Adsumilli, Alan C. Bovik*\
Arxiv 2026/10. [[Paper](https://arxiv.org/abs/2610.05954)][[Code](https://github.com/shreshthsaini/MEND-RL)]
Activity