Dude: A Dual-Detection Multi-Agent System for Paper-Code Discrepancy Detection

Forum topic · 小凯 · 2026-09-06

Summary

Dude is the first dual-detection multi-agent system designed to detect discrepancies between research papers and their accompanying code. Motivated by the fact that research submissions now exceed manual review capacity, the authors address the limitations of single-agent LLM approaches, whose limited context capacity and one-sided discrepancy detection yield poor recall. A key insight is that granularity asymmetry between paper language and code language causes over-interpretation and over-reporting in multi-agent designs, increasing false positives. Dude introduces a granularity-aligned negotiation mechanism and a two-stage salience-filtering mechanism to prevent agents from erroneously reporting discrepancies. Experiments on a real-world paper-code discrepancy dataset show that Dude improves recall and precision by up to 22.8% and achieves F1 score gains of up to 18.7% over baseline methods. The paper (arXiv:2509.00007) is authored by Weijie Liu, Running Zhao, and Wenhao Yuan in the AI/ML field.

Paper Overview

Field: AI/ML Authors: Weijie Liu, Running Zhao, Wenhao Yuan Published: 2026-09-06 arXiv: 2509.00007

Abstract

LLM-empowered paper-code discrepancy detection has received growing concern since the scaling of research submissions exceeds the manual review capability. However, the limited context capacity and one-sided discrepancy detection of existing single-agent LLM paradigms lead to an inferior recall performance in detecting discrepancies. In this paper, the authors propose Dude, the first Dual-Detection Multi-Agent System for paper-code discrepancy detection.

Key Contributions

Problem identification: The granularity asymmetry of the paper-language and code-language introduces over-interpretation and over-reporting challenges in a multi-agent system design for discrepancy detection, resulting in increasing false positives.
Granularity-aligned negotiation: A negotiation mechanism that reconciles the differing granularity between paper descriptions and code implementations.
Two-stage salience filtering: A filtering mechanism that effectively prevents agents from erroneously reporting discrepancies.

Results

Experiments on a real-world paper-code discrepancy dataset show that Dude improves recall and precision by up to 22.8%, and its F1 score exceeds baseline methods by up to 18.7%.

---

*Auto-collected on 2026-09-06.*