f-divergences

#f-divergences

$f$-Trajectory Balance: A Loss Family for Tuning GFlowNets, Generative Models, and LLMs with Off- and On-Policy Data

arXiv cs.LG ↗ · 2026-05-18 Cached

This paper introduces a family of loss functions derived from f-divergences for training generative models like GFlowNets and LLMs, which are valid off-policy while matching on-policy gradients of the corresponding f-divergence. Applications include molecule discovery and asynchronous LLM tuning.

0 favorites 0 likes

f-divergences

$f$-Trajectory Balance: A Loss Family for Tuning GFlowNets, Generative Models, and LLMs with Off- and On-Policy Data

Submit Feedback