![[2602.02050] Rethinking the Role of Entropy in Optimizing Tool-Use Behaviors for Large Language Model Agents](https://arxiv.org/static/browse/0.3.4/images/arxiv-logo-fb.png){kind=logo}
[2602.02050] Rethinking the Role of Entropy in Optimizing Tool-Use Behaviors for Large Language Model Agents
Summary
This article explores the role of entropy in optimizing tool-use behaviors for Large Language Model (LLM) agents, highlighting the correlation between entropy reduction and improved tool call quality.
Why It Matters
Understanding how entropy influences tool-use behaviors in LLMs is crucial for enhancing their efficiency and performance in real-world applications. This research provides a novel approach to managing tool calls, which can significantly reduce latency and improve overall agent adaptability.
Key Takeaways
- Entropy reduction correlates positively with high-quality tool calls.
- Two reward strategies—sparse outcome and dense process rewards—enhance tool-use behavior.
- Sparse rewards can reduce tool calls by over 72%, while dense rewards improve performance by 22%.
Computer Science > Artificial Intelligence arXiv:2602.02050 (cs) [Submitted on 2 Feb 2026 (v1), last revised 18 Feb 2026 (this version, v2)] Title:Rethinking the Role of Entropy in Optimizing Tool-Use Behaviors for Large Language Model Agents Authors:Zeping Li, Hongru Wang, Yiwen Zhao, Guanhua Chen, Yixia Li, Keyang Chen, Yixin Cao, Guangnan Ye, Hongfeng Chai, Zhenfei Yin View a PDF of the paper titled Rethinking the Role of Entropy in Optimizing Tool-Use Behaviors for Large Language Model Agents, by Zeping Li and 9 other authors View PDF HTML (experimental) Abstract:Tool-using agents based on Large Language Models (LLMs) excel in tasks such as mathematical reasoning and multi-hop question answering. However, in long trajectories, agents often trigger excessive and low-quality tool calls, increasing latency and degrading inference performance, making managing tool-use behavior challenging. In this work, we conduct entropy-based pilot experiments and observe a strong positive correlation between entropy reduction and high-quality tool calls. Building on this finding, we propose using entropy reduction as a supervisory signal and design two reward strategies to address the differing needs of optimizing tool-use behavior. Sparse outcome rewards provide coarse, trajectory-level guidance to improve efficiency, while dense process rewards offer fine-grained supervision to enhance performance. Experiments across diverse domains show that both reward designs improve tool-use behav...