TUMIX通过并行多智能体协作，提升大模型推理准确率。
TUMIX: Multi-Agent Test-Time Scaling with Tool-Use Mixture
- 多个智能体并行运行，采用不同工具使用策略
- 在多个基准上平均提升3.55%准确率，成本相近
- 支持动态停止，仅需49%计算量仍保持高性能
尽管集成代码解释器和搜索等工具已显著增强如ChatGPT Agent和Gemini-Pro等大语言模型的推理能力，但关于最优工具使用实践的指导仍然缺乏。核心挑战在于如何有效结合文本推理、编程与搜索以应对多样化问题。本文提出工具使用混合框架TUMIX，该框架并行运行多个智能体，每个智能体采用不同的工具使用策略和答案路径，并基于问题与先前回答迭代共享与优化结果。实验表明，TUMIX在多个关键推理基准上相较于最先进的工具增强与测试时扩展方法，平均准确率提升达3.55%，且推理成本相近。我们发现智能体多样性与质量至关重要，可通过大模型自动优化设计实现提升。此外，TUMIX可在达到足够置信度时提前终止精炼过程，性能损失极小，仅需49%的推理成本；进一步扩展可获得更高性能，代价也相应增加。
原文摘要 · Abstract (English)
While integrating tools like Code Interpreter and Search has significantly enhanced Large Language Model (LLM) reasoning in models like ChatGPT Agent and Gemini-Pro, practical guidance on optimal tool use is lacking. The core challenge is effectively combining textual reasoning, coding, and search for diverse questions. In this paper, we propose Tool-Use Mixture (TUMIX), an ensemble framework that runs multiple agents in parallel, each employing distinct tool-use strategies and answer paths. Agents in TUMIX iteratively share and refine responses based on the question and previous answers. In experiments, TUMIX achieves significant gains over state-of-the-art tool-augmented and test-time scaling methods, delivering an average accuracy improvement of up to 3.55% over the best baseline on Gemini-2.5-Pro and Gemini-2.5-Flash across key reasoning benchmarks, with near-equal inference costs. We find that agent diversity and quality are crucial and can be enhanced by using LLMs to auto-optimize agent designs. Furthermore, TUMIX can halt refinement upon reaching sufficient confidence, preserving performance at only 49% of the inference cost. Further scaling can achieve higher performance, albeit at a greater cost.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。