循环语言模型改进组合式工具调用

摘要

循环语言模型在推理基准上展现出令人瞩目的结果，但其在智能体工具使用方面的潜力尚未得到充分探索。我们在组合式工具调用的场景中研究该问题——在此类场景中，模型需要协调多个API调用、维护中间状态，并保持工具交互之间的依赖关系。我们在API-Bank、BFCL和NESTful上评估了原生与改造型循环语言模型，在匹配的监督微调设置下比较了循环和非循环模型，并考察了推理时不同循环深度的影响。受控实验表明，循环计算通常有助于组合式和依赖感知的工具使用，而对孤立API调用的提升较小且更依赖模型本身。多步工具使用的准确率总体上随循环深度增加而提高；然而，自适应推理通过仅在必要时分配额外计算，实现了更优的计算-性能权衡。我们的结果表明，循环语言模型是智能体系统的一种有前景的架构，尤其适用于需要可靠规划、协调与执行组合式工具使用工作流的场景。

English

Looped language models have shown promising results on reasoning benchmarks, yet their potential for agentic tool use remains largely unexplored. We study this question in compositional tool-calling settings, where models must coordinate multiple API calls, maintain intermediate state, and preserve dependencies across tool interactions. We evaluate native and retrofitted looped language models on API-Bank, BFCL, and NESTful, comparing looped and non-looped models trained under matched supervised fine-tuning recipes and varying recurrent depth at inference time. In controlled experiments, recurrent computation generally benefits compositional and dependency-aware tool use, while providing smaller and more model-dependent gains on isolated API invocation. Accuracy on multi-step tool use generally increases with recurrent depth; adaptive inference, however, achieves a more favorable compute-performance trade-off by allocating additional computation only when needed. Our results suggest that looped language models are a promising architecture for agentic systems that require reliable planning, coordination, and execution of compositional tool use workflows.