Salvato in:
Dettagli Bibliografici
Autori principali: Chen, Ruishuo, Wang, Xun, Hu, Rui, Li, Zhuoran, Huang, Longbo
Natura: Preprint
Pubblicazione: 2025
Soggetti:
Accesso online:https://arxiv.org/abs/2505.20110
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866916044989267968
author Chen, Ruishuo
Wang, Xun
Hu, Rui
Li, Zhuoran
Huang, Longbo
author_facet Chen, Ruishuo
Wang, Xun
Hu, Rui
Li, Zhuoran
Huang, Longbo
contents Generative Flow Networks (GFlowNets) excel at sampling diverse, high-reward objects. In many practical applications where active reward queries are infeasible, these models must be trained using static offline datasets. Prevailing training methods typically rely on a proxy model to provide reward feedback for online sampled trajectories. However, constructing a reliable proxy is often challenging due to data scarcity or high evaluation costs. While existing proxy-free approaches attempt to address this, they often impose coarse constraints that limit the model's ability to explore effectively. To overcome these limitations, we propose Trajectory-Distilled GFlowNet (TD-GFN), a novel proxy-free training framework. TD-GFN utilizes inverse reinforcement learning (IRL) to extract dense, transition-level edge rewards from offline trajectories, providing rich structural guidance for efficient exploration. Crucially, to ensure robustness, these rewards guide the policy indirectly through DAG pruning and prioritized backward sampling. This design ensures that gradient updates rely exclusively on ground-truth terminal rewards from the dataset, thereby preventing error propagation. Empirical results demonstrate that TD-GFN significantly outperforms a broad range of existing baselines in both convergence speed and sample quality, establishing a more robust and efficient paradigm for offline GFlowNet training.
format Preprint
id arxiv_https___arxiv_org_abs_2505_20110
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Beyond the Proxy: Trajectory-Distilled Guidance for Offline GFlowNet Training
Chen, Ruishuo
Wang, Xun
Hu, Rui
Li, Zhuoran
Huang, Longbo
Machine Learning
Artificial Intelligence
Generative Flow Networks (GFlowNets) excel at sampling diverse, high-reward objects. In many practical applications where active reward queries are infeasible, these models must be trained using static offline datasets. Prevailing training methods typically rely on a proxy model to provide reward feedback for online sampled trajectories. However, constructing a reliable proxy is often challenging due to data scarcity or high evaluation costs. While existing proxy-free approaches attempt to address this, they often impose coarse constraints that limit the model's ability to explore effectively. To overcome these limitations, we propose Trajectory-Distilled GFlowNet (TD-GFN), a novel proxy-free training framework. TD-GFN utilizes inverse reinforcement learning (IRL) to extract dense, transition-level edge rewards from offline trajectories, providing rich structural guidance for efficient exploration. Crucially, to ensure robustness, these rewards guide the policy indirectly through DAG pruning and prioritized backward sampling. This design ensures that gradient updates rely exclusively on ground-truth terminal rewards from the dataset, thereby preventing error propagation. Empirical results demonstrate that TD-GFN significantly outperforms a broad range of existing baselines in both convergence speed and sample quality, establishing a more robust and efficient paradigm for offline GFlowNet training.
title Beyond the Proxy: Trajectory-Distilled Guidance for Offline GFlowNet Training
topic Machine Learning
Artificial Intelligence
url https://arxiv.org/abs/2505.20110