MedCausalX: Adaptive Causal Reasoning with Self-Reflection for Trustworthy Medical Vision-Language Models

2026-03-24

AI Summary

This work addresses the vulnerability of existing medical vision-language models to spurious correlations and their lack of explicit causal reasoning mechanisms, which undermines clinical reliability. To this end, we propose MedCausalX—the first end-to-end causal reasoning framework with self-reflection capabilities. Our approach introduces causal and verification tokens, constructs contrastive samples distinguishing causal from spurious associations, and integrates a two-stage adaptive reflection architecture with a trajectory-level causal correction objective to enhance interpretability and trustworthiness. Evaluated on the newly curated CRMed dataset, MedCausalX achieves state-of-the-art performance across multiple benchmarks, improving diagnostic consistency by 5.4 points, reducing hallucination by over 10 points, and attaining the highest spatial localization IoU.

Technology Category

• Knowledge Representation and Reasoning: Diagnosis and Abductive Reasoning
• Reasoning under Uncertainty: Causality
• Computer Vision: Multi-modal Vision

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs

Abstract

Vision-Language Models (VLMs) have enabled interpretable medical diagnosis by integrating visual perception with linguistic reasoning. Yet, existing medical chain-of-thought (CoT) models lack explicit mechanisms to represent and enforce causal reasoning, leaving them vulnerable to spurious correlations and limiting their clinical reliability. We pinpoint three core challenges in medical CoT reasoning: how to adaptively trigger causal correction, construct high-quality causal-spurious contrastive samples, and maintain causal consistency across reasoning trajectories. To address these challenges, we propose MedCausalX, an end-to-end framework explicitly models causal reasoning chains in medical VLMs. We first introduce the CRMed dataset providing fine-grained anatomical annotations, structured causal reasoning chains, and counterfactual variants that guide the learning of causal relationships beyond superficial correlations. Building upon CRMed, MedCausalX employs a two-stage adaptive reflection architecture equipped with $\langle$causal$\rangle$ and $\langle$verify$\rangle$ tokens, enabling the model to autonomously determine when and how to perform causal analysis and verification. Finally, a trajectory-level causal correction objective optimized through error-attributed reinforcement learning refines the reasoning chain, allowing the model to distinguish genuine causal dependencies from shortcut associations. Extensive experiments on multiple benchmarks show that MedCausalX consistently outperforms state-of-the-art methods, improving diagnostic consistency by +5.4 points, reducing hallucination by over 10 points, and attaining top spatial grounding IoU, thereby setting a new standard for causally grounded medical reasoning.

Problem

• causal reasoning
• medical vision-language models
• spurious correlations
• chain-of-thought
• clinical reliability

Innovation

• causal reasoning
• self-reflection
• medical vision-language models
• counterfactual augmentation
• reinforcement learning