可适配且精准：企业场景 LLM Function-Calling 能力训练 Pipeline
Adaptable and Precise: Enterprise-Scenario LLM Function-Calling Capability Training Pipeline
AI总结：
针对企业场景中通用 LLM function-calling 能力不足的问题，提出包含数据合成增强、模型微调与评估的训练 pipeline，基于 Qwen2.5-Coder-7B-Instruct 和 LoRA 微调的模型在测试集准确率上超越 GPT-4 和 GPT-4o。
摘要：
企业拥有散布于各种功能中的海量 API 资产，构成了现有业务流程的骨干。通过将这些 API 作为功能工具，企业可以设计由本地化 function-calling 模型作为核心引擎驱动的、多样化的场景特定智能体应用。然而，通用模型在计算效率、输出准确性和稳定性方面往往无法满足企业需求，因此需要进行场景特定的适配。在本文中，我们提出了一种针对真实业务场景量身定制的 function-calling 能力训练 pipeline。该 pipeline 包括场景特定 function-calling 数据的合成与增强、模型微调以及性能评估与分析。使用该 pipeline，我们在数字 HR agent 场景中生成了 1,260 个完全由 AI 生成的样本和 1,035 个增强的人工标注样本。采用 Qwen2.5-Coder-7B-Instruct 模型作为基础模型，并在四块具有 24GB VRAM 的 GPU 上使用 LoRA 方法进行微调。我们微调后的模型在评估和实际应用中表现出色，在测试集上的准确率超越了 GPT-4 和 GPT-4o。这些结果验证了所提出的 pipeline 在训练场景特定 function-calling 模型方面的可靠性。
英文摘要：
Enterprises possess a vast array of API assets scattered across various functions, forming the backbone of existing business processes. By leveraging these APIs as functional tools, enterprises can design diverse, scenario-specific agent applications, driven by on-premise function-calling models as the core engine. However, generic models often fail to meet enterprise requirements in terms of computational efficiency, output accuracy, and stability, necessitating scenario-specific adaptation. In this paper, we propose a training pipeline for function-calling capabilities tailored to real-world business scenarios. This pipeline includes the synthesis and augmentation of scenario-specific function-calling data, model fine-tuning, and performance evaluation and analysis. Using this pipeline, we generated 1,260 fully AI-generated samples and 1,035 augmented manually-labeled samples in digital HR agent scenario. The Qwen2.5-Coder-7B-Instruct model was employed as the base model and fine-tuned using the LoRA method on four GPUs with 24GB VRAM. Our fine-tuned model demonstrated outstanding performance in evaluations and practical applications, surpassing GPT-4 and GPT-4o in accuracy on the test set. These results validate the reliability of the proposed pipeline for training scenario-specific function-calling models.
机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。
