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 2503.04725cs.CLcs.AIcs.ITcs.LGmath.ITphysics.data-an
收藏

L$^2$M:长上下文语言建模的互信息缩放定律

L$^2$M: Mutual Information Scaling Law for Long-Context Language Modeling

NSF AI Institute for Artificial Intelligence and Fundamental Interactions(NSF人工智能与基础相互作用研究院);Massachusetts Institute of Technology(麻省理工学院);Polytechnic University of Catalonia(加泰罗尼亚理工大学);Harvard University(哈佛大学);University of California, Los Angeles(加州大学洛杉矶分校)

Zhuo Chen, Oriol Mayné i Comas, Zhuotao Jin, Di Luo, Marin Soljačić

AI总结:

本文提出基于二分互信息缩放定律的通用理论框架,推导长上下文语言建模条件,并在Transformer和状态空间模型上验证,为高效长上下文架构设计提供原则性基础。

摘要:

我们提出了一个用于理解长上下文语言建模的通用理论框架,该框架基于一个在自然语言中严格验证的二分互信息缩放定律。我们证明,二分互信息捕获了与常规两点互信息不同且独立缩放的多词元交互,并表明这为准确建模长序列所需的依赖关系提供了更完整的刻画。利用这一缩放定律,我们提出了长上下文语言建模(L$^2$M)条件,该条件为模型的历史状态(即负责存储过去信息的潜在变量)的必要缩放设定了下界,以实现有效的长上下文建模。我们在Transformer和状态空间模型上验证了该框架及其预测。我们的工作为理解长上下文建模以及设计具有更强长上下文能力的更高效架构提供了原则性基础,并具有超越自然语言的潜在应用。

英文摘要:

We present a universal theoretical framework for understanding long-context language modeling based on a bipartite mutual information scaling law that we rigorously verify in natural language. We demonstrate that bipartite mutual information captures multi-token interactions distinct from and scaling independently of conventional two-point mutual information, and show that this provides a more complete characterization of the dependencies needed for accurately modeling long sequences. Leveraging this scaling law, we formulate the Long-context Language Modeling (L$^2$M) condition, which lower bounds the necessary scaling of a model's history state -- the latent variables responsible for storing past information -- for effective long-context modeling. We validate the framework and its predictions on transformer and state-space models. Our work provides a principled foundation to understand long-context modeling and to design more efficient architectures with stronger long-context capabilities, with potential applications beyond natural language.

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。

↑