Conservative Offline Goal-Conditioned Implicit V-Learning

Kaiqiang Ke, Qian Lin, Zongkai Liu, Shenghong He, Chao Yu
Proceedings of the 42nd International Conference on Machine Learning, PMLR 267:29591-29607, 2025.

Abstract

Offline goal-conditioned reinforcement learning (GCRL) learns a goal-conditioned value function to train policies for diverse goals with pre-collected datasets. Hindsight experience replay addresses the issue of sparse rewards by treating intermediate states as goals but fails to complete goal-stitching tasks where achieving goals requires stitching different trajectories. While cross-trajectory sampling is a potential solution that associates states and goals belonging to different trajectories, we demonstrate that this direct method degrades performance in goal-conditioned tasks due to the overestimation of values on unconnected pairs. To this end, we propose Conservative Goal-Conditioned Implicit Value Learning (CGCIVL), a novel algorithm that introduces a penalty term to penalize value estimation for unconnected state-goal pairs and leverages the quasimetric framework to accurately estimate values for connected pairs. Evaluations on OGBench, a benchmark for offline GCRL, demonstrate that CGCIVL consistently surpasses state-of-the-art methods across diverse tasks.

Cite this Paper

BibTeX
@InProceedings{pmlr-v267-ke25a, title = {Conservative Offline Goal-Conditioned Implicit V-Learning}, author = {Ke, Kaiqiang and Lin, Qian and Liu, Zongkai and He, Shenghong and Yu, Chao}, booktitle = {Proceedings of the 42nd International Conference on Machine Learning}, pages = {29591--29607}, year = {2025}, editor = {Singh, Aarti and Fazel, Maryam and Hsu, Daniel and Lacoste-Julien, Simon and Berkenkamp, Felix and Maharaj, Tegan and Wagstaff, Kiri and Zhu, Jerry}, volume = {267}, series = {Proceedings of Machine Learning Research}, month = {13--19 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v267/main/assets/ke25a/ke25a.pdf}, url = {https://proceedings.mlr.press/v267/ke25a.html}, abstract = {Offline goal-conditioned reinforcement learning (GCRL) learns a goal-conditioned value function to train policies for diverse goals with pre-collected datasets. Hindsight experience replay addresses the issue of sparse rewards by treating intermediate states as goals but fails to complete goal-stitching tasks where achieving goals requires stitching different trajectories. While cross-trajectory sampling is a potential solution that associates states and goals belonging to different trajectories, we demonstrate that this direct method degrades performance in goal-conditioned tasks due to the overestimation of values on unconnected pairs. To this end, we propose Conservative Goal-Conditioned Implicit Value Learning (CGCIVL), a novel algorithm that introduces a penalty term to penalize value estimation for unconnected state-goal pairs and leverages the quasimetric framework to accurately estimate values for connected pairs. Evaluations on OGBench, a benchmark for offline GCRL, demonstrate that CGCIVL consistently surpasses state-of-the-art methods across diverse tasks.} }
Endnote
%0 Conference Paper %T Conservative Offline Goal-Conditioned Implicit V-Learning %A Kaiqiang Ke %A Qian Lin %A Zongkai Liu %A Shenghong He %A Chao Yu %B Proceedings of the 42nd International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2025 %E Aarti Singh %E Maryam Fazel %E Daniel Hsu %E Simon Lacoste-Julien %E Felix Berkenkamp %E Tegan Maharaj %E Kiri Wagstaff %E Jerry Zhu %F pmlr-v267-ke25a %I PMLR %P 29591--29607 %U https://proceedings.mlr.press/v267/ke25a.html %V 267 %X Offline goal-conditioned reinforcement learning (GCRL) learns a goal-conditioned value function to train policies for diverse goals with pre-collected datasets. Hindsight experience replay addresses the issue of sparse rewards by treating intermediate states as goals but fails to complete goal-stitching tasks where achieving goals requires stitching different trajectories. While cross-trajectory sampling is a potential solution that associates states and goals belonging to different trajectories, we demonstrate that this direct method degrades performance in goal-conditioned tasks due to the overestimation of values on unconnected pairs. To this end, we propose Conservative Goal-Conditioned Implicit Value Learning (CGCIVL), a novel algorithm that introduces a penalty term to penalize value estimation for unconnected state-goal pairs and leverages the quasimetric framework to accurately estimate values for connected pairs. Evaluations on OGBench, a benchmark for offline GCRL, demonstrate that CGCIVL consistently surpasses state-of-the-art methods across diverse tasks.
APA
Ke, K., Lin, Q., Liu, Z., He, S. & Yu, C.. (2025). Conservative Offline Goal-Conditioned Implicit V-Learning. Proceedings of the 42nd International Conference on Machine Learning, in Proceedings of Machine Learning Research 267:29591-29607 Available from https://proceedings.mlr.press/v267/ke25a.html.

Related Material