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 built to detect discrepancies between research papers and their accompanying code, addressing limitations of single-agent LLM approaches. As research submissions grow faster than manual review capacity, LLM-based discrepancy detection has gained attention, but existing single-agent paradigms suffer from limited context capacity and one-sided analysis, resulting in poor recall. The authors identify a granularity asymmetry between paper language and code language that introduces over-interpretation and over-reporting challenges in multi-agent designs, inflating false positives. Dude counters this with a granularity-aligned negotiation mechanism and a two-stage salience-filtering process that prevents agents from misreporting discrepancies. Experiments on real paper-code discrepancy datasets show Dude improves recall and accuracy by up to 22.8% and achieves F1 scores up to 18.7% higher than 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

Summary

LLM-empowered paper-code discrepancy detection has received growing attention since the scaling of research submissions exceeds manual review capability. However, the limited context capacity and one-sided discrepancy detection of existing single-agent LLM paradigms lead to 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. They discover that the granularity asymmetry between paper-language and code-language introduces over-interpretation and over-reporting challenges in multi-agent system designs for discrepancy detection, resulting in increased false positives.

To address this, Dude introduces:

Granularity-aligned negotiation: aligning the differing granularity of paper and code languages across agents to avoid misinterpretation.
Two-stage salience-filtering mechanism: effectively prevents agents from wrongly reporting discrepancies.

Results

Experiments on real-world paper-code discrepancy datasets show that Dude significantly improves recall and accuracy by up to 22.8%, and achieves F1 scores up to 18.7% higher than baseline methods.

---

*Auto-collected on 2026-09-06. Source: zhichai.net forum post.*

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

Summary

Paper Overview

Summary

Results

Tags