Research

Paper

AI LLM February 26, 2026

Exploratory Memory-Augmented LLM Agent via Hybrid On- and Off-Policy Optimization

Authors

Zeyuan Liu, Jeonghye Kim, Xufang Luo, Dongsheng Li, Yuqing Yang

Abstract

Exploration remains the key bottleneck for large language model agents trained with reinforcement learning. While prior methods exploit pretrained knowledge, they fail in environments requiring the discovery of novel states. We propose Exploratory Memory-Augmented On- and Off-Policy Optimization (EMPO$^2$), a hybrid RL framework that leverages memory for exploration and combines on- and off-policy updates to make LLMs perform well with memory while also ensuring robustness without it. On ScienceWorld and WebShop, EMPO$^2$ achieves 128.6% and 11.3% improvements over GRPO, respectively. Moreover, in out-of-distribution tests, EMPO$^2$ demonstrates superior adaptability to new tasks, requiring only a few trials with memory and no parameter updates. These results highlight EMPO$^2$ as a promising framework for building more exploratory and generalizable LLM-based agents.

Metadata

arXiv ID: 2602.23008
Provider: ARXIV
Primary Category: cs.LG
Published: 2026-02-26
Fetched: 2026-02-27 04:35

Related papers

Raw Data (Debug)
{
  "raw_xml": "<entry>\n    <id>http://arxiv.org/abs/2602.23008v1</id>\n    <title>Exploratory Memory-Augmented LLM Agent via Hybrid On- and Off-Policy Optimization</title>\n    <updated>2026-02-26T13:50:57Z</updated>\n    <link href='https://arxiv.org/abs/2602.23008v1' rel='alternate' type='text/html'/>\n    <link href='https://arxiv.org/pdf/2602.23008v1' rel='related' title='pdf' type='application/pdf'/>\n    <summary>Exploration remains the key bottleneck for large language model agents trained with reinforcement learning. While prior methods exploit pretrained knowledge, they fail in environments requiring the discovery of novel states. We propose Exploratory Memory-Augmented On- and Off-Policy Optimization (EMPO$^2$), a hybrid RL framework that leverages memory for exploration and combines on- and off-policy updates to make LLMs perform well with memory while also ensuring robustness without it. On ScienceWorld and WebShop, EMPO$^2$ achieves 128.6% and 11.3% improvements over GRPO, respectively. Moreover, in out-of-distribution tests, EMPO$^2$ demonstrates superior adaptability to new tasks, requiring only a few trials with memory and no parameter updates. These results highlight EMPO$^2$ as a promising framework for building more exploratory and generalizable LLM-based agents.</summary>\n    <category scheme='http://arxiv.org/schemas/atom' term='cs.LG'/>\n    <category scheme='http://arxiv.org/schemas/atom' term='cs.AI'/>\n    <published>2026-02-26T13:50:57Z</published>\n    <arxiv:comment>Accepted to ICLR 2026</arxiv:comment>\n    <arxiv:primary_category term='cs.LG'/>\n    <author>\n      <name>Zeyuan Liu</name>\n    </author>\n    <author>\n      <name>Jeonghye Kim</name>\n    </author>\n    <author>\n      <name>Xufang Luo</name>\n    </author>\n    <author>\n      <name>Dongsheng Li</name>\n    </author>\n    <author>\n      <name>Yuqing Yang</name>\n    </author>\n  </entry>"
}