Learn-to-Distance: Distance Learning for Detecting LLM-Generated Text

2026-01-29

AI Summary

This work proposes a geometry-aware detection method based on a rewriting mechanism to address the risks of misinformation and academic integrity violations posed by large language model (LLM)-generated text. The approach introduces adaptive distance learning into a rewriting-based detection framework for the first time, dynamically learning the geometric distance between original and rewritten texts to significantly enhance detection performance. Theoretical analysis demonstrates the superiority of this adaptive strategy over fixed-distance metrics and elucidates its generalization mechanism. Extensive experiments across more than 100 configurations involving multiple mainstream LLMs show that the proposed method substantially outperforms existing baselines, achieving relative performance improvements ranging from 57.8% to 80.6% across different target models.

Technology Category

• Natural Language Processing: Fact-Checking / Misinformation Detection (NLP Focus)
• Machine Learning: Large Multimodal Models (LMMs)
• Search and Optimization: Learning to Search

Application Category

• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Search and Retrieval-Augmented AI: Search Tool Learning with LLM: Teaching LLMs to invoke search and make use of retrieved information
• User Modeling, Personalization and Recommendation: Large Language Models (LLM) for user modeling and recommendation

Abstract

Modern large language models (LLMs) such as GPT, Claude, and Gemini have transformed the way we learn, work, and communicate. Yet, their ability to produce highly human-like text raises serious concerns about misinformation and academic integrity, making it an urgent need for reliable algorithms to detect LLM-generated content. In this paper, we start by presenting a geometric approach to demystify rewrite-based detection algorithms, revealing their underlying rationale and demonstrating their generalization ability. Building on this insight, we introduce a novel rewrite-based detection algorithm that adaptively learns the distance between the original and rewritten text. Theoretically, we demonstrate that employing an adaptively learned distance function is more effective for detection than using a fixed distance. Empirically, we conduct extensive experiments with over 100 settings, and find that our approach demonstrates superior performance over baseline algorithms in the majority of scenarios. In particular, it achieves relative improvements from 57.8\% to 80.6\% over the strongest baseline across different target LLMs (e.g., GPT, Claude, and Gemini).

Problem

• LLM-generated text detection
• distance learning
• text authenticity
• misinformation
• academic integrity

Innovation

• distance learning
• LLM-generated text detection
• adaptive distance function
• rewrite-based detection
• geometric analysis