arXiv 2601.036762026-01-07cs.CLcs.AI

通过技能分类引导的数据合成实现大语言模型的组合泛化

Towards Compositional Generalization of LLMs via Skill Taxonomy Guided Data Synthesis

State Key Laboratory of Complex & Critical Software Environment, Beihang University（复杂与关键软件环境国家重点实验室，北京航空航天大学）；Beijing Academy of Artificial Intelligence（北京人工智能研究院）；Beijing Institute of Technology（北京理工大学）；Institute of Computing Technology, CAS（中国科学院计算技术研究所）

Yifan Wei, Li Du, Xiaoyan Yu, Yang Feng, Angsheng Li

备注：: The code and data for our methods and experiments are available at https://github.com/weiyifan1023/STEPS

阅读 PDF

arXiv 页面 ↗HTML 原文 ↗

AI总结：

STEPS通过技能分类引导的数据合成提升大语言模型的组合泛化能力。

摘要：

大型语言模型（LLMs）和基于代理的系统往往由于数据瓶颈而难以实现组合泛化，这种瓶颈在于复杂技能组合遵循长尾、幂律分布，限制了指令遵循性能和代理中心任务的泛化能力。为了解决这一挑战，我们提出了STEPS，一种由技能分类引导的熵基后训练数据合成框架，用于生成具有组合挑战性的数据。STEPS通过揭示技能间的潜在关系，并利用结构性信息理论将它们组织成可解释的、层次化的技能分类体系。基于此分类，我们将数据合成建模为一个受约束的信息最大化问题，选择在层次结构内最大化边际结构性信息的同时保持语义一致性。在具有挑战性的指令遵循基准测试中，STEPS在现有数据合成基线之上表现更优，同时在下游基于代理的评估中也实现了改进的组合泛化能力。

英文摘要：

Large Language Models (LLMs) and agent-based systems often struggle with compositional generalization due to a data bottleneck in which complex skill combinations follow a long-tailed, power-law distribution, limiting both instruction-following performance and generalization in agent-centric tasks. To address this challenge, we propose STEPS, a Skill Taxonomy guided Entropy-based Post-training data Synthesis framework for generating compositionally challenging data. STEPS explicitly targets compositional generalization by uncovering latent relationships among skills and organizing them into an interpretable, hierarchical skill taxonomy using structural information theory. Building on this taxonomy, we formulate data synthesis as a constrained information maximization problem, selecting skill combinations that maximize marginal structural information within the hierarchy while preserving semantic coherence. Experiments on challenging instruction-following benchmarks show that STEPS outperforms existing data synthesis baselines, while also yielding improved compositional generalization in downstream agent-based evaluations.

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。