mlsanityData to 7 Oct 2026

A Benchmark for Hallucination Detection in VLMs for Gastrointestinal Endoscopy

arXiv 2606.24115 ↗First posted 23 Jun 2026cs.CV cs.AIcited by 0

Not on the map. Why?

Abstract (arXiv)

Vision-language models (VLMs) are prone to hallucination, which remains a major barrier to their safe deployment in clinical practice. To date, most hallucination detection methods have been evaluated on radiology benchmarks such as MIMIC-CXR and VQA-RAD, while gastrointestinal (GI) endoscopy remains largely underexplored. In this paper, we benchmark nine hallucination detection methods on the Gut-VLM dataset, a GI diagnostic Visual Question Answering (VQA) dataset with 4,392 test VQA pairs, across five VLMs (MedGemma-4B, MedGemma-27B, LLaVA-Med-7B, LLaVA-v1.6-7B, and Lingshu-32B). The methods span three categories: black-box methods (RadFlag, SelfCheckGPT-NLI), gray-box methods (AvgProb, AvgEnt, MaxProb, MaxEnt, Semantic Entropy, and VASE), and a white-box method (ReXTrust). Our results show that ReXTrust, a white-box method, achieves the highest AUC across all five models, outperforming the strongest alternative method on each VLM by a statistically significant margin (paired permutation test, p < 0.001 in all cases), reaching a peak AUC of 93.0 on MedGemma-4B. White-box hidden-state access provides a consistent advantage of 19.5 AUC points on average (range: 9.5--33.5), with ReXTrust maintaining strong performance even on LLaVA-v1.6-7B (AUC 79.9), where black-box methods and clustering-based gray-box methods collapse to near-chance performance. Among non-white-box methods, token-level gray-box statistics (MaxEnt, MaxProb) are the strongest alternatives, outperforming both clustering-based gray-box methods (Semantic Entropy, VASE) and black-box approaches on average. We further identify confident confabulation, a failure mode in which models hallucinate with high inter-sample consistency or high token-level probability, as a systemic failure for both consistency and uncertainty-based methods.

Key facts. No citing papers in the map yet (data through 7 Oct 2026). Builds on 18 of its 25 references that are in the map. First posted on arXiv 23 Jun 2026.

Builds on

18 of 25 references.

References · 18

  1. 2023Visual Instruction Tuningcited by 7,792
  2. 2023LLaVA-Med: Training a Large Language-and-Vision Assistant for Biomedicine in One Daycited by 1,287
  3. 2021DeBERTaV3: Improving DeBERTa using ELECTRA-Style Pre-Training with Gradient-Disentangled Embedding Sharingcited by 1,219
  4. 2023SelfCheckGPT: Zero-Resource Black-Box Hallucination Detection for Generative Large Language Modelscited by 787
  5. 2023The Internal State of an LLM Knows When It's Lyingcited by 644
  6. 2019MIMIC-CXR-JPG, a large publicly available database of labeled chest radiographscited by 590
  7. 2020PathVQA: 30000+ Questions for Medical Visual Question Answeringcited by 467
  8. 2021SLAKE: A Semantically-Labeled Knowledge-Enhanced Dataset for Medical Visual Question Answeringcited by 444
  9. 2025MedGemma Technical Reportcited by 398
  10. 2025Lingshu: A Generalist Foundation Model for Unified Multimodal Medical Understanding and Reasoningcited by 217
  11. 2024GREEN: Generative Radiology Report Evaluation and Error Notationcited by 124
  12. 2024A Survey on Uncertainty Quantification of Large Language Models: Taxonomy, Open Research Challenges, and Future Directionscited by 115
6 more
  1. 2024Reference-free Hallucination Detection for Large Vision-Language Modelscited by 23
  2. 2024RadFlag: A Black-Box Hallucination Detection Method for Medical Vision Language Modelscited by 16
  3. 2025FactSelfCheck: Fact-Level Black-Box Hallucination Detection for LLMscited by 14
  4. 2024DHCP: Detecting Hallucinations by Cross-modal Attention Pattern in Large Vision-Language Modelscited by 11
  5. 2024ReXTrust: A Model for Fine-Grained Hallucination Detection in AI-Generated Radiology Reportscited by 11
  6. 2025Hallucination-Aware Multimodal Benchmark for Gastrointestinal Image Analysis with Large Vision-Language Modelscited by 9

Built on by

Not cited yet.

Similar, not cited

Close in wording, no citation either way

  1. 2024Detecting and Evaluating Medical Hallucinations in Large Vision Language Models
  2. 2026HalluCXR: Benchmarking and Mitigating Hallucinations in Medical Vision-Language Models for Chest Radiograph Interpretation
  3. 2026Med-StepBench: A Hierarchical Reasoning Framework for Evaluating Hallucinations in Medical Vision-Language Models
  4. 2025UniVRSE: Unified Vision-conditioned Response Semantic Entropy for Hallucination Detection in Medical Vision-Language Models
  5. 2024MedVH: Towards Systematic Evaluation of Hallucination for Large Vision Language Models in the Medical Context
  6. 2026Deterministic Hallucination Detection in Medical VQA via Confidence-Evidence Bayesian Gain
  7. 2026VIHD: Visual Intervention-based Hallucination Detection for Medical Visual Question Answering
  8. 2025MedHallTune: An Instruction-Tuning Benchmark for Mitigating Medical Hallucination in Vision-Language Models