Salvato in:
Dettagli Bibliografici
Autori principali: Opryshko, Evgenii, Quan, Junwei, Voelcker, Claas, Du, Yilun, Gilitschenski, Igor
Natura: Preprint
Pubblicazione: 2025
Soggetti:
Accesso online:https://arxiv.org/abs/2510.07257
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866911709615095808
author Opryshko, Evgenii
Quan, Junwei
Voelcker, Claas
Du, Yilun
Gilitschenski, Igor
author_facet Opryshko, Evgenii
Quan, Junwei
Voelcker, Claas
Du, Yilun
Gilitschenski, Igor
contents Offline goal-conditioned reinforcement learning (GCRL) often struggles with long-horizon tasks, where errors in value estimation accumulate and produce unreliable policies. It is typically assumed that effective long-term planning is infeasible without specialized training. In contrast, our work demonstrates that existing GCRL policies can complete long-horizon tasks when combined with a lightweight, training-free planning wrapper. We find that standard goal-conditioned value functions encode locally consistent geometric structure sufficient for planning. Our approach, Test-Time Graph Search (TTGS), constructs a graph over the offline dataset and employs an adaptive subgoal selection strategy. To address unreliable value estimates during shortest-path search, we propose a novel mechanism that softly penalizes long-distance transitions. Our method incurs negligible computational overhead and requires no additional supervision or parameter updates. On the OGBench benchmark, TTGS significantly boosts success rates across multiple base learners and tasks, with primary gains on challenging long-horizon locomotion tasks where some success rates are improved from near-zero to over 90\%, often matching or outperforming methods that require complex auxiliary training. Code and videos can be found at https://ktolnos.github.io/ttgs.
format Preprint
id arxiv_https___arxiv_org_abs_2510_07257
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Test-Time Graph Search for Goal-Conditioned Reinforcement Learning
Opryshko, Evgenii
Quan, Junwei
Voelcker, Claas
Du, Yilun
Gilitschenski, Igor
Machine Learning
Offline goal-conditioned reinforcement learning (GCRL) often struggles with long-horizon tasks, where errors in value estimation accumulate and produce unreliable policies. It is typically assumed that effective long-term planning is infeasible without specialized training. In contrast, our work demonstrates that existing GCRL policies can complete long-horizon tasks when combined with a lightweight, training-free planning wrapper. We find that standard goal-conditioned value functions encode locally consistent geometric structure sufficient for planning. Our approach, Test-Time Graph Search (TTGS), constructs a graph over the offline dataset and employs an adaptive subgoal selection strategy. To address unreliable value estimates during shortest-path search, we propose a novel mechanism that softly penalizes long-distance transitions. Our method incurs negligible computational overhead and requires no additional supervision or parameter updates. On the OGBench benchmark, TTGS significantly boosts success rates across multiple base learners and tasks, with primary gains on challenging long-horizon locomotion tasks where some success rates are improved from near-zero to over 90\%, often matching or outperforming methods that require complex auxiliary training. Code and videos can be found at https://ktolnos.github.io/ttgs.
title Test-Time Graph Search for Goal-Conditioned Reinforcement Learning
topic Machine Learning
url https://arxiv.org/abs/2510.07257