Meta-RL Induces Exploration in Language Agents

TL;DR

This paper introduces LaMer, a Meta-RL framework that enhances language agents' exploration and adaptation capabilities, leading to significant performance improvements and better generalization in complex, unseen tasks.

Contribution

LaMer is a novel Meta-RL approach that enables language agents to actively explore and adapt without gradient updates, improving robustness and performance.

Findings

11-19% performance gains on benchmark tasks

Better generalization to unseen environments

Effective in diverse multi-turn tasks

Abstract

Reinforcement learning (RL) has enabled the training of large language model (LLM) agents to interact with the environment and to solve multi-turn long-horizon tasks. However, the RL-trained agents often struggle in tasks that require active exploration and fail to efficiently adapt from trial-and-error experiences. In this paper, we present LaMer, a general Meta-RL framework that enables LLM agents to actively explore and learn from the environment feedback at test time. LaMer consists of two key components: (i) a cross-episode training framework to encourage exploration and long-term rewards optimization; and (ii) in-context policy adaptation via reflection, allowing the agent to adapt their policy from task feedback signal without gradient update. Experiments across diverse environments show that LaMer significantly improves performance over RL baselines, with 11%, 14%, and 19% performance gains on Sokoban, MineSweeper and Webshop, respectively. Moreover, LaMer also demonstrates better generalization to more challenging or previously unseen tasks compared to the RL-trained agents. Overall, our results demonstrate that Meta-RL provides a principled approach to induce exploration in language agents, enabling more robust adaptation to novel environments through learned exploration strategies.