A Study on Leveraging Search and Self-Feedback for Agent Reasoning

TL;DR

This paper investigates how search algorithms and self-feedback mechanisms can enhance reasoning in language agents, highlighting the importance of ground-truth or well-designed feedback for effective reasoning across different tasks.

Contribution

It provides a comparative analysis of ground-truth and self-feedback during search and proposes task-specific feedback strategies to improve reasoning in complex tasks.

Findings

Self-feedback can limit generalization in reasoning tasks.

Ground-truth feedback significantly improves search effectiveness.

Task-specific feedback mechanisms are necessary for complex reasoning.

Abstract

Recent works have demonstrated that incorporating search during inference can significantly improve reasoning capabilities of language agents. Some approaches may make use of the ground truth or rely on model's own generated feedback. The search algorithm uses this feedback to then produce values that will update its criterion for exploring and exploiting various reasoning paths. In this study, we investigate how search and model's self-feedback can be leveraged for reasoning tasks. First, we explore differences in ground-truth feedback and self-feedback during search for math reasoning. Second, we observe limitations in applying search techniques to more complex tasks like tool-calling and design domain-specific approaches to address these gaps. Our experiments reveal challenges related to generalization when solely relying on self-feedback during search. For search to work effectively, either access to the ground-truth is needed or feedback mechanisms need to be carefully designed for the specific task.