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 2602.11044cs.CLcs.AI
收藏

通过端到端微分实现语言模型逆向

Language Model Inversion through End-to-End Differentiation

School of Informatics, University of Edinburgh, Edinburgh, UK(信息学院,爱丁堡大学,爱丁堡,英国)

Kevin Yandoka Denamganaï, Kartic Subr

AI总结:

本研究提出通过端到端微分实现语言模型的逆向,利用梯度下降优化提示以生成目标输出,验证了方法在不同长度提示和目标上的有效性。

摘要:

尽管涌现出大量关于语言模型(LM)的研究,但很少有方法分析LM的可逆性。也就是说,给定一个LM和一个期望的目标输出序列,确定哪些输入提示能产生目标输出仍然是一个开放性问题。我们把这个问题 formulation 为一个经典的基于梯度的优化问题。首先,我们提出了一种简单的算法,以实现给定(冻结)LM的端到端可微性,然后通过梯度下降找到优化的提示。我们的核心见解是将LM视为在token分布序列上操作的函数(而不是传统上视为在token序列上的函数)。我们的实验和消融研究证明,我们的DLM驱动的逆向能够可靠且高效地优化长度为10和80的提示,以目标长度为20的几个白盒LM(即开箱即用)。\n

英文摘要:

Despite emerging research on Language Models (LM), few approaches analyse the invertibility of LMs. That is, given a LM and a desirable target output sequence of tokens, determining what input prompts would yield the target output remains an open problem. We formulate this problem as a classical gradient-based optimisation. First, we propose a simple algorithm to achieve end-to-end differentiability of a given (frozen) LM and then find optimised prompts via gradient descent. Our central insight is to view LMs as functions operating on sequences of distributions over tokens (rather than the traditional view as functions on sequences of tokens). Our experiments and ablations demonstrate that our DLM-powered inversion can reliably and efficiently optimise prompts of lengths $10$ and $80$ for targets of length $20$, for several white-box LMs (out-of-the-box).

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。

↑