Scaling Test-time Compute for LLM Agents

TL;DR

This paper systematically explores how applying various test-time scaling strategies can enhance the effectiveness of large language model agents, focusing on methods like parallel sampling, revision, verification, and diversification.

Contribution

It introduces and evaluates multiple test-time scaling strategies for language agents, providing insights into their impact on performance and guiding future improvements.

Findings

Scaling test-time compute improves agent performance.

Reflecting at appropriate times is crucial for effectiveness.

List-wise verification and merging outperform other methods.

Abstract

Scaling test time compute has shown remarkable success in improving the reasoning abilities of large language models (LLMs). In this work, we conduct the first systematic exploration of applying test-time scaling methods to language agents and investigate the extent to which it improves their effectiveness. Specifically, we explore different test-time scaling strategies, including: (1) parallel sampling algorithms; (2) sequential revision strategies; (3) verifiers and merging methods; (4)strategies for diversifying rollouts.We carefully analyze and ablate the impact of different design strategies on applying test-time scaling on language agents, and have follow findings: 1. Scaling test time compute could improve the performance of agents. 2. Knowing when to reflect is important for agents. 3. Among different verification and result merging approaches, the list-wise method performs best. 4. Increasing diversified rollouts exerts a positive effect on the agent's task performance.