Runaway is Ashamed, But Helpful: On the Early-Exit Behavior of Large Language Model-based Agents in Embodied Environments

TL;DR

This paper investigates early-exit strategies for large language model-based agents in embodied environments, aiming to improve efficiency by reducing redundant steps with minimal performance loss.

Contribution

It introduces intrinsic and extrinsic early-exit methods for LLM agents, along with evaluation metrics, demonstrating significant efficiency gains across multiple models and environments.

Findings

Early-exit methods reduce redundant steps significantly.

Minor performance degradation observed with early-exit strategies.

Assistance from stronger agents post early-exit improves overall performance.

Abstract

Agents powered by large language models (LLMs) have demonstrated strong planning and decision-making capabilities in complex embodied environments. However, such agents often suffer from inefficiencies in multi-turn interactions, frequently trapped in repetitive loops or issuing ineffective commands, leading to redundant computational overhead. Instead of relying solely on learning from trajectories, we take a first step toward exploring the early-exit behavior for LLM-based agents. We propose two complementary approaches: 1. an $\textbf{intrinsic}$ method that injects exit instructions during generation, and 2. an $\textbf{extrinsic}$ method that verifies task completion to determine when to halt an agent's trial. To evaluate early-exit mechanisms, we introduce two metrics: one measures the reduction of $\textbf{redundant steps}$ as a positive effect, and the other evaluates $\textbf{progress degradation}$ as a negative effect. Experiments with 4 different LLMs across 5 embodied environments show significant efficiency improvements, with only minor drops in agent performance. We also validate a practical strategy where a stronger agent assists after an early-exit agent, achieving better performance with the same total steps. We will release our code to support further research.