arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2602.11871cs.CLcs.LG
收藏

DMAP:文本的分布图

DMAP: A Distribution Map for Text

University of Manchester, UK(曼彻斯特大学,英国);Featurespace, a Visa Solution(Visa解决方案的Featurespace);Risk and Security AI Lab, Visa Inc., UK(Visa公司的风险与安全AI实验室,英国);University of Cambridge, UK(剑桥大学,英国)

Tom Kempton, Julia Rozanova, Parameswaran Kamalaruban, Maeve Madigan, Karolina Wresilo, Yoann L. Launay, David Sutton, Stuart Burrell

AI总结:

DMAP通过语言模型将文本映射到单位区间内的样本集,编码排名和概率信息,用于高效分析文本并支持多种应用,包括生成参数验证和机器生成文本检测。

摘要:

大型语言模型(LLMs)是统计文本分析的强大工具,其派生的下一个令牌概率分布序列提供了丰富的信息。提取此信号通常依赖于如困惑度等指标,但这些指标未能充分考虑上下文;如何解释给定下一个令牌的概率取决于条件分布形状中编码的合理选择数量。本文提出DMAP,一种数学上严谨的方法,通过语言模型将文本映射到单位区间内的样本集,共同编码排名和概率信息。这种表示使高效、模型无关的分析成为可能,并支持多种应用。我们通过三个案例研究展示了其用途:(i)验证生成参数以确保数据完整性,(ii)考察概率曲率在机器生成文本检测中的作用,(iii)对下游模型进行法医分析,揭示在合成数据后训练留下的统计指纹。我们的结果表明,DMAP提供了一种统一的文本统计视图,计算简单,可在消费级硬件上运行,具有广泛的应用性,并为利用LLMs进行文本分析的进一步研究奠定了基础。

英文摘要:

Large Language Models (LLMs) are a powerful tool for statistical text analysis, with derived sequences of next-token probability distributions offering a wealth of information. Extracting this signal typically relies on metrics such as perplexity, which do not adequately account for context; how one should interpret a given next-token probability is dependent on the number of reasonable choices encoded by the shape of the conditional distribution. In this work, we present DMAP, a mathematically grounded method that maps a text, via a language model, to a set of samples in the unit interval that jointly encode rank and probability information. This representation enables efficient, model-agnostic analysis and supports a range of applications. We illustrate its utility through three case studies: (i) validation of generation parameters to ensure data integrity, (ii) examining the role of probability curvature in machine-generated text detection, and (iii) a forensic analysis revealing statistical fingerprints left in downstream models that have been subject to post-training on synthetic data. Our results demonstrate that DMAP offers a unified statistical view of text that is simple to compute on consumer hardware, widely applicable, and provides a foundation for further research into text analysis with LLMs.

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。

↑