Saved in:
Bibliographic Details
Main Authors: Gu, Yuliang, Cao, Hongpeng, Caccamo, Marco, Hovakimyan, Naira
Format: Preprint
Published: 2025
Subjects:
Online Access:https://arxiv.org/abs/2507.19437
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866918104409309184
author Gu, Yuliang
Cao, Hongpeng
Caccamo, Marco
Hovakimyan, Naira
author_facet Gu, Yuliang
Cao, Hongpeng
Caccamo, Marco
Hovakimyan, Naira
contents Capturing latent variations ("contexts") is key to deploying reinforcement-learning (RL) agents beyond their training regime. We recast context-based RL as a dual inference-control problem and formally characterize two properties and their hierarchy: observation sufficiency (preserving all predictive information) and control sufficiency (retaining decision-making relevant information). Exploiting this dichotomy, we derive a contextual evidence lower bound(ELBO)-style objective that cleanly separates representation learning from policy learning and optimizes it with Bottlenecked Contextual Policy Optimization (BCPO), an algorithm that places a variational information-bottleneck encoder in front of any off-policy policy learner. On standard continuous-control benchmarks with shifting physical parameters, BCPO matches or surpasses other baselines while using fewer samples and retaining performance far outside the training regime. The framework unifies theory, diagnostics, and practice for context-based RL.
format Preprint
id arxiv_https___arxiv_org_abs_2507_19437
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Observations Meet Actions: Learning Control-Sufficient Representations for Robust Policy Generalization
Gu, Yuliang
Cao, Hongpeng
Caccamo, Marco
Hovakimyan, Naira
Machine Learning
Capturing latent variations ("contexts") is key to deploying reinforcement-learning (RL) agents beyond their training regime. We recast context-based RL as a dual inference-control problem and formally characterize two properties and their hierarchy: observation sufficiency (preserving all predictive information) and control sufficiency (retaining decision-making relevant information). Exploiting this dichotomy, we derive a contextual evidence lower bound(ELBO)-style objective that cleanly separates representation learning from policy learning and optimizes it with Bottlenecked Contextual Policy Optimization (BCPO), an algorithm that places a variational information-bottleneck encoder in front of any off-policy policy learner. On standard continuous-control benchmarks with shifting physical parameters, BCPO matches or surpasses other baselines while using fewer samples and retaining performance far outside the training regime. The framework unifies theory, diagnostics, and practice for context-based RL.
title Observations Meet Actions: Learning Control-Sufficient Representations for Robust Policy Generalization
topic Machine Learning
url https://arxiv.org/abs/2507.19437