让大模型自己回忆过往成功经验，更好用工具完成复杂任务
Self-Guided Function Calling in Large Language Models via Stepwise Experience Recall
- 通过逐步回溯过往成功操作路径，动态更新经验池
- 在ToolQA上易题提升6.1%，难题提升4.7%；τ-bench上准确率最高提升23.38%
- 适合需要长期积累工具使用经验的复杂任务场景
函数调用使大语言模型能通过工具和API与外部系统交互。面对多步工具使用任务时，模型仍难以正确选择工具、生成参数及规划工具链。现有方法通常依赖人工设计的任务特异性示例或从精心整理的库中检索，需大量专家投入，且随着工具多样性和任务难度增加，提示工程变得愈发复杂低效。为此，我们提出自引导方法Stepwise Experience Recall（SEER），通过细粒度、逐步方式从持续更新的经验池中检索。不同于静态或人工维护的库，SEER将过去成功的执行轨迹不断注入经验池，实现池子的持续扩展与模型性能的渐进提升。在ToolQA基准测试中，SEER在简单问题上平均提升6.1%，在困难问题上提升4.7%。进一步在包含两个真实场景的τ-bench上测试，基于Qwen2.5-7B与Qwen2.5-72B模型，分别取得7.44%与23.38%的显著准确率提升。
原文摘要 · Abstract (English)
Function calling enables large language models (LLMs) to interact with external systems by leveraging tools and APIs. When faced with multi-step tool usage, LLMs still struggle with tool selection, parameter generation, and tool-chain planning. Existing methods typically rely on manually designing task-specific demonstrations, or retrieving from a curated library. These approaches demand substantial expert effort and prompt engineering becomes increasingly complex and inefficient as tool diversity and task difficulty scale. To address these challenges, we propose a self-guided method, Stepwise Experience Recall (SEER), which performs fine-grained, stepwise retrieval from a continually updated experience pool. Instead of relying on static or manually curated library, SEER incrementally augments the experience pool with past successful trajectories, enabling continuous expansion of the pool and improved model performance over time. Evaluated on the ToolQA benchmark, SEER achieves an average improvement of 6.1% on easy and 4.7% on hard questions. We further test SEER on $τ$-bench, which includes two real-world domains. Powered by Qwen2.5-7B and Qwen2.5-72B models, SEER demonstrates substantial accuracy gains of 7.44% and 23.38%, respectively.
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。