TRACE：基于业务规则的推理课程——面向企业级大语言模型中知识保持的参数化工具检索

摘要

参数化检索通过为每个API分配唯一的虚拟令牌，并训练模型通过受限集束搜索生成这些令牌，使大语言模型能够隐式检索工具。工具感知指出该机制存在两个关键缺陷：训练过程会破坏参数化工具知识，且其集束搜索解码速度过慢，难以满足实时部署需求。我们提出TRACE（基于增强思维链与企业规则的工具检索），这是一种通过两阶段课程学习解决该分离问题的方法。第一阶段复用工具感知中的多格式记忆化监督微调，利用低秩适配植入工具知识。第二阶段是我们的核心贡献：模型在生成JSON格式工具令牌列表之前，需要输出思维追踪过程，训练数据包含两类来源——工具感知中的RRB对，以及针对领域专家整理的业务规则所合成的查询——均附有推理追踪标记。该训练目标在保持第一阶段多项选择提问和问答探测准确率的同时，可实现生产级延迟下的单光束贪婪解码。在涵盖两个企业产品线共8300+工具的企业级工具库评估中，TRACE第二阶段训练不仅保留了工具理解能力，更实现了性能提升：与第一阶段相比，多项选择提问准确率提升3.2个百分点，问答探测准确率提升9个百分点。在检索任务中，TRACE在领域A上达到约86%的召回率，在领域B上达到约60%——相比之下，嵌入基线仅有约27%和52%——均采用单光束贪婪解码，可直接部署于生产环境。

English

Parametric retrieval enables LLMs to retrieve tools implicitly by assigning each API a unique virtual token and training the model to generate it via constrained beam search. Toolsense shows that this regime has two critical drawbacks: it destroys parametric tool knowledge during training, and its beam-search decoding is too slow for real-time deployment. We introduce TRACE (Tool Retrieval via Augmented Chain-of-thought and Enterprise rules), a two-stage curriculum that resolves this dissociation. Stage 1 reuses the multi-format memorization SFT from ToolSense to seed tool knowledge with LoRA. Stage 2 is our core contribution: the model is trained to emit a thinking trace before producing a JSON list of tool tokens, using two data sources -- RRB pairs from ToolSense and queries synthesized to target business rules curated by domain experts -- both augmented with reasoning traces. This training objective preserves Stage 1 MCQ and QA probing accuracy while enabling single-beam greedy decoding at production latency. Evaluated on a combined enterprise catalog of 8,300+ tools across two enterprise product lines, TRACE training for Stage 2 not only preserves but improves tool understanding: MCQ accuracy gains +3.2 pp and QA probing gains +9 pp over Stage 1. On retrieval, TRACE achieves ~86% recall on Domain A and ~60% on Domain B -- compared to embedding baseline performance of ~27% & ~52% -- both with single-beam greedy decoding, making it directly deployable at production latency.