Look Before You Leap: Autonomous Exploration for LLM Agents

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Ye, Ziang, Shi, Wentao, Liu, Yuxin, Wang, Yu, Cai, Zhengzhou, Shi, Yaorui, Gu, Qi, Cai, Xunliang, Feng, Fuli
Format: Preprint
Published: 2026
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866918504408547328
author Ye, Ziang
Shi, Wentao
Liu, Yuxin
Wang, Yu
Cai, Zhengzhou
Shi, Yaorui
Gu, Qi
Cai, Xunliang
Feng, Fuli
author_facet Ye, Ziang
Shi, Wentao
Liu, Yuxin
Wang, Yu
Cai, Zhengzhou
Shi, Yaorui
Gu, Qi
Cai, Xunliang
Feng, Fuli
contents Large language model based agents often fail in unfamiliar environments due to premature exploitation: a tendency to act on prior knowledge before acquiring sufficient environment-specific information. We identify autonomous exploration as a critical yet underexplored capability for building adaptive agents. To formalize and quantify this capability, we introduce Exploration Checkpoint Coverage, a verifiable metric that measures how broadly an agent discovers key states, objects, and affordances. Our systematic evaluation reveals that agents trained with standard task-oriented reinforcement learning consistently exhibit narrow and repetitive behaviors that impede downstream performance. To address this limitation, we develop a training strategy that interleaves task-execution rollouts and exploration rollouts, with each type of rollout optimized by its corresponding verifiable reward. Building on this training strategy, we propose the Explore-then-Act paradigm, which decouples information-gathering from task execution: agents first utilize an interaction budget to acquire grounded environmental knowledge, then leverage it for task resolution. Our results demonstrate that learning to systematically explore is imperative for building generalizable and real-world-ready agents.
format Preprint
id arxiv_https___arxiv_org_abs_2605_16143
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle Look Before You Leap: Autonomous Exploration for LLM Agents
Ye, Ziang
Shi, Wentao
Liu, Yuxin
Wang, Yu
Cai, Zhengzhou
Shi, Yaorui
Gu, Qi
Cai, Xunliang
Feng, Fuli
Artificial Intelligence
Computation and Language
Large language model based agents often fail in unfamiliar environments due to premature exploitation: a tendency to act on prior knowledge before acquiring sufficient environment-specific information. We identify autonomous exploration as a critical yet underexplored capability for building adaptive agents. To formalize and quantify this capability, we introduce Exploration Checkpoint Coverage, a verifiable metric that measures how broadly an agent discovers key states, objects, and affordances. Our systematic evaluation reveals that agents trained with standard task-oriented reinforcement learning consistently exhibit narrow and repetitive behaviors that impede downstream performance. To address this limitation, we develop a training strategy that interleaves task-execution rollouts and exploration rollouts, with each type of rollout optimized by its corresponding verifiable reward. Building on this training strategy, we propose the Explore-then-Act paradigm, which decouples information-gathering from task execution: agents first utilize an interaction budget to acquire grounded environmental knowledge, then leverage it for task resolution. Our results demonstrate that learning to systematically explore is imperative for building generalizable and real-world-ready agents.
title Look Before You Leap: Autonomous Exploration for LLM Agents
topic Artificial Intelligence
Computation and Language
url https://arxiv.org/abs/2605.16143