How Can Input Reformulation Improve Tool Usage Accuracy in a Complex Dynamic Environment? A Study on $τ$-bench

2025-08-28

AI Summary

Large language models (LLMs) exhibit inconsistent reasoning, strategy drift, and inaccurate long-range information extraction in multi-turn dynamic tool-use scenarios (e.g., τ-bench). To address these challenges, we propose the Input-Reformulation Multi-Agent (IRMA) framework. IRMA automatically reformulates user queries to explicitly integrate domain-specific rules and tool-call suggestions, while leveraging multi-agent coordination to enhance decision consistency. Methodologically, it unifies ReAct prompting, native function calling, and self-reflection capabilities, augmented by human-in-the-loop error analysis and iterative policy refinement. Experimental results demonstrate that IRMA achieves absolute improvements of 16.1%, 12.7%, and 19.1% over ReAct, Function Calling, and Self-Reflection baselines, respectively, on the pass⁵ metric. This yields substantial gains in reliability, focus, and behavioral stability during complex interactive tool use—establishing a novel paradigm for controllable, adaptive tool orchestration in dynamic environments.

Technology Category

• Machine Learning: Large Multimodal Models (LMMs)
• Multiagent Systems: Multiagent Learning
• Humans and AI: Human-Aware Planning and Behavior Prediction

Application Category

• Search and Retrieval-Augmented AI: Search Tool Learning with LLM: Teaching LLMs to invoke search and make use of retrieved information
• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions
• User Modeling, Personalization and Recommendation: Large Language Models (LLM) for user modeling and recommendation

Abstract

Recent advances in reasoning and planning capabilities of large language models (LLMs) have enabled their potential as autonomous agents capable of tool use in dynamic environments. However, in multi-turn conversational environments like $τ$-bench, these agents often struggle with consistent reasoning, adherence to domain-specific policies, and extracting correct information over a long horizon of tool-calls and conversation. To capture and mitigate these failures, we conduct a comprehensive manual analysis of the common errors occurring in the conversation trajectories. We then experiment with reformulations of inputs to the tool-calling agent for improvement in agent decision making. Finally, we propose the Input-Reformulation Multi-Agent (IRMA) framework, which automatically reformulates user queries augmented with relevant domain rules and tool suggestions for the tool-calling agent to focus on. The results show that IRMA significantly outperforms ReAct, Function Calling, and Self-Reflection by 16.1%, 12.7%, and 19.1%, respectively, in overall pass^5 scores. These findings highlight the superior reliability and consistency of IRMA compared to other methods in dynamic environments.

Problem

• Improving tool usage accuracy in dynamic environments
• Addressing consistent reasoning in multi-turn conversations
• Enhancing agent decision making with input reformulation

Innovation

• Input-Reformulation Multi-Agent framework
• Augments queries with domain rules
• Automatically suggests tools for agents