Saved in:
Bibliographic Details
Main Authors: Lei, Zengxiang, Shreekumar, Ananth, Rosenthal, Jonathan, Song, Ruoyu, Cardenas, Alvaro A., Fremont, Daniel J., Xu, Dongyan, Ukkusuri, Satish, Celik, Z. Berkay
Format: Preprint
Published: 2026
Subjects:
Online Access:https://arxiv.org/abs/2605.01729
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866915975725580288
author Lei, Zengxiang
Shreekumar, Ananth
Rosenthal, Jonathan
Song, Ruoyu
Cardenas, Alvaro A.
Fremont, Daniel J.
Xu, Dongyan
Ukkusuri, Satish
Celik, Z. Berkay
author_facet Lei, Zengxiang
Shreekumar, Ananth
Rosenthal, Jonathan
Song, Ruoyu
Cardenas, Alvaro A.
Fremont, Daniel J.
Xu, Dongyan
Ukkusuri, Satish
Celik, Z. Berkay
contents Generative Flow Networks (GFlowNets) learn to sample states proportional to an unnormalized reward. Despite their theoretical promise, practical training is often unstable, exhibiting severe loss spikes and mode collapse. To tackle this, we first assess the sensitivity of GFlowNet objectives, demonstrating that a small Total Variation (TV) distance between the learned and target distributions does not preclude unbounded training loss. Motivated by this mismatch, we establish converse guarantees by deriving loss-to-TV bounds that certify global fidelity from bounded trajectory balance losses. Lastly, we propose Stable GFlowNets, an algorithm that leverages our theoretical results to stabilize training, and empirically demonstrate improved training behavior and superior distributional fidelity.
format Preprint
id arxiv_https___arxiv_org_abs_2605_01729
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle Stable GFlowNets with Probabilistic Guarantees
Lei, Zengxiang
Shreekumar, Ananth
Rosenthal, Jonathan
Song, Ruoyu
Cardenas, Alvaro A.
Fremont, Daniel J.
Xu, Dongyan
Ukkusuri, Satish
Celik, Z. Berkay
Machine Learning
Generative Flow Networks (GFlowNets) learn to sample states proportional to an unnormalized reward. Despite their theoretical promise, practical training is often unstable, exhibiting severe loss spikes and mode collapse. To tackle this, we first assess the sensitivity of GFlowNet objectives, demonstrating that a small Total Variation (TV) distance between the learned and target distributions does not preclude unbounded training loss. Motivated by this mismatch, we establish converse guarantees by deriving loss-to-TV bounds that certify global fidelity from bounded trajectory balance losses. Lastly, we propose Stable GFlowNets, an algorithm that leverages our theoretical results to stabilize training, and empirically demonstrate improved training behavior and superior distributional fidelity.
title Stable GFlowNets with Probabilistic Guarantees
topic Machine Learning
url https://arxiv.org/abs/2605.01729