返回全部动态
LLM令牌生成的动力系统可区分性保证
原标题：Guarantees on Dynamical System Distinguishability for LLM Token Generation
AI 摘要
该论文将LLM响应分类任务形式化为两个随机线性动力系统之间的二元假设检验，证明忽略令牌动力学的分类器存在准确率下限，而基于动力系统的分类误分类概率随序列长度指数衰减。研究还通过近似交织条件刻画了跨嵌入模型的泛化能力，解释了该方法的经验性能。
以上摘要由 AI 生成，可能存在误差。事实请以原文为准。
正文节选
Computer Science > Machine Learning Title:Guarantees on Dynamical System Distinguishability for LLM Token Generation View PDF HTML (experimental) Abstract:Recent work has shown that classifying large language models (LLMs)' responses can be distinguished by modeling token embeddings as trajectories of a black-box dynamical system (DS) and comparing prediction residuals of two DSs. Despite the empirical success of this dynamical approach, a theoretical understanding of why it works, h
发布时间：2026-08-03 12:00
抓取时间：2026-08-03 15:21
来源机构：arXiv