XrayClaw: Cooperative-Competitive Multi-Agent Alignment for Trustworthy Chest X-ray Diagnosis

arXiv cs.CV / 4/6/2026

📰 NewsSignals & Early TrendsIdeas & Deep AnalysisModels & Research

共有:

Key Points

The paper presents XrayClaw, a cooperative–competitive multi-agent framework aimed at making chest X-ray (CXR) AI diagnoses more trustworthy and logically consistent.
It uses four specialized cooperative agents to mirror a systematic clinical workflow, plus a competitive auditor agent to independently verify outputs.
To reduce reasoning hallucinations and consensus failures common in monolithic or single-model multi-agent setups, the authors introduce Competitive Preference Optimization that penalizes illogical reasoning and enforces mutual verification.
Experiments on MS-CXR-T, MIMIC-CXR, and CheXbench show state-of-the-art gains in diagnostic accuracy, clinical reasoning fidelity, and zero-shot domain generalization.
The work positions cooperative–competitive multi-agent alignment as a new paradigm for reliable automated medical imaging analysis.

Abstract

Chest X-ray (CXR) interpretation is a fundamental yet complex clinical task that increasingly relies on artificial intelligence for automation. However, traditional monolithic models often lack the nuanced reasoning required for trustworthy diagnosis, frequently leading to logical inconsistencies and diagnostic hallucinations. While multi-agent systems offer a potential solution by simulating collaborative consultations, existing frameworks remain susceptible to consensus-based errors when instantiated by a single underlying model. This paper introduces XrayClaw, a novel framework that operationalizes multi-agent alignment through a sophisticated cooperative-competitive architecture. XrayClaw integrates four specialized cooperative agents to simulate a systematic clinical workflow, alongside a competitive agent that serves as an independent auditor. To reconcile these distinct diagnostic pathways, we propose Competitive Preference Optimization, a learning objective that penalizes illogical reasoning by enforcing mutual verification between analytical and holistic interpretations. Extensive empirical evaluations on the MS-CXR-T, MIMIC-CXR, and CheXbench benchmarks demonstrate that XrayClaw achieves state-of-the-art performance in diagnostic accuracy, clinical reasoning fidelity, and zero-shot domain generalization. Our results indicate that XrayClaw effectively mitigates cumulative hallucinations and enhances the overall reliability of automated CXR diagnosis, establishing a new paradigm for trustworthy medical imaging analysis.