DMAP: A Distribution Map for Text

TL;DR

DMAP is a novel, mathematically grounded method that maps text via language models into a set of samples encoding rank and probability, enabling versatile, efficient analysis of generated and real text.

Contribution

Introduces DMAP, a model-agnostic, mathematically grounded approach for analyzing text with language models, supporting diverse applications like data integrity and forensic analysis.

Findings

DMAP provides a unified statistical view of text.

It is simple to compute on consumer hardware.

Demonstrates utility in validation, detection, and forensic analysis.

Abstract

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.