arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~
arXiv 2502.06589cs.CLcs.AIcs.LG
收藏

Hephaestus:通过持续预训练提升大型语言模型的基础智能体能力

Hephaestus: Improving Fundamental Agent Capabilities of Large Language Models through Continual Pre-Training

Georgia Institute of Technology(佐治亚理工学院);Amazon(亚马逊公司)
Yuchen Zhuang, Jingfeng Yang, Haoming Jiang, Xin Liu, Kewei Cheng, Sanket Lokegaonkar, Yifan Gao, Qing Ping, Tianyi Liu, Binxuan Huang, Zheng Li, Zhengyang Wang… 展开作者

Yuchen Zhuang, Jingfeng Yang, Haoming Jiang, Xin Liu, Kewei Cheng, Sanket Lokegaonkar, Yifan Gao, Qing Ping, Tianyi Liu, Binxuan Huang, Zheng Li, Zhengyang Wang, Pei Chen, Ruijie Wang, Rongzhi Zhang, Nasser Zalmout, Priyanka Nigam, Bing Yin, Chao Zhang

AI总结:

针对面向智能体的预训练数据稀缺问题,提出首个大规模语料库Hephaestus-Forge(含103B智能体专用数据),通过持续预训练提升LLM智能体的API调用、推理规划等基础能力,在三个基准上优于中小开源LLM且媲美商业LLM。

摘要:

由于面向智能体的预训练数据稀缺,基于LLM的自主智能体通常依赖复杂提示或大量微调,但这些方法往往难以在保留强泛化性的同时引入新能力。我们提出Hephaestus-Forge,这是首个大规模预训练语料库,旨在增强LLM智能体在API函数调用、内在推理与规划,以及适应环境反馈方面的基础能力。Hephaestus-Forge包含103B智能体专用数据,涵盖76,537个API,既包括用于介绍API函数知识的工具文档,也包括用于强化内在推理的函数调用轨迹。为探索有效训练协议,我们研究了缩放定律以确定数据混合比的最优方案。通过在Hephaestus-Forge上持续预训练,Hephaestus在三个智能体基准上优于中小规模开源LLM,并可与商业LLM媲美,证明了我们的预训练语料库在增强LLM的基础智能体能力及对新任务或环境的泛化性方面的有效性。

英文摘要:

Due to the scarcity of agent-oriented pre-training data, LLM-based autonomous agents typically rely on complex prompting or extensive fine-tuning, which often fails to introduce new capabilities while preserving strong generalizability. We introduce Hephaestus-Forge, the first large-scale pre-training corpus designed to enhance the fundamental capabilities of LLM agents in API function calling, intrinsic reasoning and planning, and adapting to environmental feedback. Hephaestus-Forge comprises 103B agent-specific data encompassing 76,537 APIs, including both tool documentation to introduce knowledge of API functions and function calling trajectories to strengthen intrinsic reasoning. To explore effective training protocols, we investigate scaling laws to identify the optimal recipe in data mixing ratios. By continual pre-training on Hephaestus-Forge, Hephaestus outperforms small- to medium-scale open-source LLMs and rivals commercial LLMs on three agent benchmarks, demonstrating the effectiveness of our pre-training corpus in enhancing fundamental agentic capabilities and generalization of LLMs to new tasks or environments.

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。

↑