Deterministic Hallucination Detection in Medical VQA via Confidence-Evidence Bayesian Gain

2026-03-23

AI Summary

This work addresses the challenge of hallucinated responses in multimodal large language models (MLLMs) for medical visual question answering (VQA), where generated answers often contradict visual evidence. Existing hallucination detection methods suffer from high computational overhead and reliance on external models. To overcome these limitations, the authors propose CEBaG—a deterministic, end-to-end hallucination detection framework that requires no sampling, external dependencies, or task-specific hyperparameters. CEBaG uniquely leverages internal model log-probabilities, integrating token-level prediction variance with visual evidence magnitude to formulate a Confidence-Evidence Bayesian Gain metric. Evaluated across four medical MLLMs and three VQA benchmarks under 16 distinct settings, CEBaG achieves the highest AUC in 13 cases, outperforming the prior state-of-the-art method VASE by an average of 8 AUC points.

Technology Category

• Machine Learning: Calibration & Uncertainty Quantification
• Computer Vision: Multi-modal Vision
• Natural Language Processing: Question Answering

Application Category

• Search and Retrieval-Augmented AI: Retrieval-Augmented Generation (RAG) and multi-modal RAG
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs

Abstract

Multimodal large language models (MLLMs) have shown strong potential for medical Visual Question Answering (VQA), yet they remain prone to hallucinations, defined as generating responses that contradict the input image, posing serious risks in clinical settings. Current hallucination detection methods, such as Semantic Entropy (SE) and Vision-Amplified Semantic Entropy (VASE), require 10 to 20 stochastic generations per sample together with an external natural language inference model for semantic clustering, making them computationally expensive and difficult to deploy in practice. We observe that hallucinated responses exhibit a distinctive signature directly in the model's own log-probabilities: inconsistent token-level confidence and weak sensitivity to visual evidence. Based on this observation, we propose Confidence-Evidence Bayesian Gain (CEBaG), a deterministic hallucination detection method that requires no stochastic sampling, no external models, and no task-specific hyperparameters. CEBaG combines two complementary signals: token-level predictive variance, which captures inconsistent confidence across response tokens, and evidence magnitude, which measures how much the image shifts per-token predictions relative to text-only inference. Evaluated across four medical MLLMs and three VQA benchmarks (16 experimental settings), CEBaG achieves the highest AUC in 13 of 16 settings and improves over VASE by 8 AUC points on average, while being fully deterministic and self-contained. The code will be made available upon acceptance.

Problem

• hallucination detection
• medical VQA
• multimodal large language models
• deterministic detection
• clinical safety

Innovation

• hallucination detection
• medical VQA
• deterministic method
• multimodal LLMs
• Bayesian gain