Medical VLMs

Vision-language models have demonstrated impressive capabilities in general medical visual question answering, yet due to limited interpretability, it remains unclear whether their predictions reflect evidence-grounded clinical reasoning or reliance on spurious priors. We introduce Med-R2 Bench, a hierarchical benchmark aligned with the clinical workflow to evaluate adversarial robustness with visual grounding. We design stepwise QA tasks to assess whether reasoning chains are strictly grounded in visual evidence across the four clinical stages, and employ adversarial perturbations to test robustness against misleading cues. Med-R2 comprises 42,432 images, 31 task categories, and 110,406 QA pairs. Evaluation across 14 VLMs reveals a sequential performance degradation along the four-stage clinical workflow. Adversarial experiments show that models rely heavily on correct prompts to guess answers. Even when provided with explicit visual cues, the models struggle to accurately align textual descriptions. Finally, we demonstrate stepwise fine-tuning using our hierarchical data significantly improves reasoning robustness, highlighting its potential to drive future improvements in evidence-based medical AI.

Effective pre-polysomnography screening for obstructive sleep apnea-hypopnea syndrome (OSAHS) requires combining clinical risk factors with visible craniofacial and neck cues. Directly prompting general-purpose multimodal foundation models for medical yes/no decisions can yield unstable, poorly calibrated outputs. We propose EviOSAHS, an evidence-grounded multimodal reasoning framework that separates image-only anatomical evidence acquisition from final clinical adjudication. Each frontal facial image is decomposed into seven fixed anatomical queries covering the neck, chin, mouth, face/neck fat, lower jaw, midface, and nose. Visual responses are converted into structured evidence cards recording target anatomy, visibility, risk direction, evidence strength, confidence, and a concise summary. These cards are combined with a cleaned clinical profile only in the final stage, where a large language model performs balanced binary screening adjudication. We evaluated EviOSAHS on a 642-subject cohort, mapping normal subjects to screening-negative and mild, moderate, or severe OSAHS subjects to screening-positive. EviOSAHS achieved 88.47% accuracy, 94.86% sensitivity, 93.74% F1-score, and a 5.14% false-negative rate, outperforming clinical-only prompting, direct multimodal prompting, and naive two-stage pipelines under a unified protocol. Ablations showed that seven-question visual decomposition and balanced final adjudication were critical to the high-sensitivity operating point. A question-level audit of 4,494 visual outputs showed a 100% structured parse rate and 93.88% high-visibility rate. EviOSAHS provides an auditable, high-sensitivity workflow for binary pre-polysomnography OSAHS screening, but should be viewed as a triage assistant rather than a diagnostic system. Prospective validation, external testing, and calibrated operating-point control are needed before clinical deployment.

CT report generation (CTRG) requires models to summarize three-dimensional anatomical context and pathological findings from hundreds of axial slices. Existing methods typically learn a direct image-to-text mapping, providing limited mechanisms for modeling how CT evidence evolves across slices or how reports respond to controlled changes in latent lesion-related factors. We propose SliceWorld, a CT-specific world-state framework that treats an axial CT scan as an ordered sequence along the z-axis. SliceWorld encodes prefix CT evidence into factor-aware latent states containing anatomy, lesion, and uncertainty components, and projects these states into world tokens used for multi-step future-slice feature prediction, lesion-factor intervention, and LLM-based report generation. The model is first pretrained on CT slice sequences with predictive, factor-aware, and counterfactual objectives, and is then fine-tuned on paired CT-report data. Experiments on M3D-Cap and CT-RATE show that SliceWorld improves natural language generation metrics and clinically oriented automatic evaluation. Further analyses demonstrate multi-horizon future-slice prediction, measurable factor alignment, reduced-slice robustness, and selective lesion-sensitive report modulation.

We introduce JMed48k, a multi-profession Japanese healthcare licensing benchmark for evaluating vision-language models. Built from official PDF materials released by the Japanese Ministry of Health, Labour and Welfare, JMed48k contains 48,862 exam questions and 20,142 images from 11 national licensing examinations between 2005 and 2025, with visual content annotated under an 8-type taxonomy. From this corpus, we derive JMed48k-Eval, a recent five-year evaluation subset with 12,484 scored questions, including 9,905 text-only questions and 2,579 questions with images. We evaluate 21 proprietary, open-source, and medical-specific models, reporting text-only and with-image performance separately. Because these subsets contain different questions, we further introduce a paired image-removal audit that evaluates questions with images before and after removing visual content to explore four answer-transition states. The audit shows that proprietary and open source models gain substantially from images, whereas medical-specific systems show limited observable use of visual evidence, with many correct answers persisting after image removal. Even among proprietary models, the net image-removal effect varies sevenfold across professions, from +5.7 points on Physician questions to +39.8 points on Public Health Nurse questions. We release JMed48k to support reproducible, profession-stratified evaluation of vision-language models in medical licensing settings.

Computed tomography (CT) is a central to three-dimensional medical imaging, yet CT-based artificial intelligence remains fragmented across task-specific models for segmentation, classification, registration, and report analysis. Here we present FlexiCT, a family of CT foundation models trained by agglomerative continual pretraining on 266,227 CT volumes from 56 publicly available datasets, forming a large-scale public resource for CT representation learning. FlexiCT uses agglomerative pretraining across three stages: two-dimensional axial pretraining, three-dimensional anatomical pretraining and report-guided semantic alignment. This training strategy supports slice-level, volume-level and vision-language analysis. Across five downstream task families (segmentation, classification, registration, vision-language understanding and clinical retrieval), FlexiCT matches or exceeds prior task-specific approaches on multiple benchmarks. Its embeddings further organize CT scans along gradients associated with various tumor stages, suggesting that CT foundation models can capture imaging features relevant to disease phenotype characterization. Project page and code are available at: https://ricklisz.github.io/flexict.github.io and https://github.com/ricklisz/FlexiCT.

Understanding surgical workflow in real time is fundamental for intelligent surgical embodiment, where AI systems continuously perceive and respond as surgery proceeds. In the operating room, critical decisions depend on subtle, moment-to-moment changes, such as fine instrument movements and evolving tissue states, where even slight perceptual delays can limit assistance or compromise safety. Yet existing methods remain offline or operate at coarse temporal scales, generating descriptions only after processing clips, preventing immediate reaction. We address this by proposing SurgOnAir, a streaming vision-language model that processes frames sequentially without future access and progressively generates narration tokens as visual input arrives. SurgOnAir achieves fine-grained frame-to-token generation, enabling instant responsiveness to evolving surgical dynamics. Built upon our curated hierarchical dataset SurgOnAir-11k spanning action-, step-, and phase-level supervision, the model is trained to produce multi-level textual responses that reflect the inherent hierarchy of surgical procedures. Furthermore, special transition tokens are generated to explicitly mark state changes, allowing SurgOnAir to capture and signal key workflow transitions as they occur. Experiments show that SurgOnAir enables real-time understanding through a single vision-language model that unifies streaming across multiple hierarchies of the surgical workflow, generating superior and hierarchy-aware narrations. Code and dataset will be public.

While medical Multimodal Large Language Models (MLLMs) have shown promise in assisting diagnosis, they still frequently generate hallucinated responses that appear linguistically plausible but lack visual evidence. Such hallucinations pose risks to clinical decision-making and necessitate effective detection. Existing introspective detection methods primarily perform uncertainty estimation or logical verification by analyzing model responses conditioned on original or perturbed inputs. However, such external perturbations are often heuristic and context-agnostic, which overlooks the internal cross-modal dependency between generated tokens and related visual tokens during decoding. To address this issue, we propose VIHD, a Visual Intervention-based Hallucination Detection method that leverages targeted visual token masking to calibrate semantic entropy for more effective hallucination detection. VIHD locates visually dominant decoder layers via Visual Dependency Probing (VDP), executes Visual Intervention Decoding (VID) via token masking to calibrate the semantic distribution, and quantifies the resulting Calibrated Semantic Entropy (CSE) as a reliable hallucination signal. Extensive experiments on three medical VQA benchmarks with two medical MLLMs demonstrate that VIHD consistently outperforms state-of-the-art methods, underscoring the importance of fine-grained visual dependency for hallucination detection. The code will be available at https://github.com/Jiayi-Chen-AU/VIHD

Experienced physicians develop diagnostic expertise through clinical practice, acquiring not only disease knowledge but also the ability to differentiate confusable conditions. Current medical vision-language models (VLMs) lack this capability -- their parameters encode static knowledge that does not evolve across diagnostic encounters. We propose MedExpMem, an experience memory framework enabling VLM-based diagnostic agents to accumulate differential diagnosis expertise. Unlike retrieval-augmented generation, which retrieves encyclopedic disease descriptions, MedExpMem memorizes discriminative experience derived from the agent's own diagnostic failures and organizes them as pairwise differential notes encoding key discriminators, actionable decision rules and reasoning error patterns. The framework adopts a two-phase construction process mirroring physician learning: initial practice exposes knowledge gaps, and reflective re-diagnosis refines understanding. When encountering new cases, the agent retrieves experience memory to guide differential reasoning. We evaluate MedExpMem on a radiology benchmark spanning 11 subspecialties. Results demonstrate consistent accuracy improvements, maximum 7.0%, across diverse models and scales. Analytical experiments validate experience quality and robustness, demonstrating MedExpMem as a competitive method addresses medical adaptation needs beyond the reach of parameteric learning.

Vision-language models (VLMs) are increasingly used for medical image interpretation, yet they frequently hallucinate, generating clinically plausible but factually incorrect findings that pose direct patient safety risks. We introduce HalluCXR, a benchmark evaluating six architecturally diverse VLMs across 856 stratified MIMIC-CXR chest radiographs and three query types, yielding 15,408 model evaluations. An eight-category hallucination taxonomy with clinical severity ratings and a two-layer detection pipeline are validated against 250 human annotations (auto-detection F1=0.959; LLM judge F1=0.907). We find that 61.9--82.3% of outputs contain hallucinations, with clinically dangerous errors in up to 80.2%. Three key patterns emerge: normal radiographs paradoxically attract the most severe hallucinations, common findings are systematically over-fabricated while rare findings go under-detected, and response length alone predicts hallucination risk (AUC up to 0.908). A six-model ensemble reduces fabrication by up to 84.8% at the cost of increased omission; a three-model subset retains comparable performance at half the cost. These results establish that hallucination auditing, verbosity-based risk monitoring, and ensemble-based safety layers are prerequisites for clinical deployment.

Large Vision Language Models (LVLMs) show promise in medical applications, but their inability to faithfully ground responses in visual evidence raises serious concerns about clinical trustworthiness. While visual attribution methods are widely used to explain LVLM predictions, whether these explanations actually reflect the visual evidence underlying the model's decision is largely unverified, since ground-truth annotations for internal model reasoning are typically unavailable. We address this question for chest X-ray (CXR) reasoning by developing a causal evaluation framework that retains only CXR-VQA samples for which the expert-annotated region is verified, via counterfactual editing, to be causally responsible for the model's prediction. Using this framework across 11 attribution methods, six open-source LVLMs, and two output modes (direct answer and step-by-step reasoning), we find that existing attribution methods often fail to identify the evidence used by LVLMs. To address this failure, we propose MedFocus, a concept-based attribution method that localizes clinically meaningful anatomical regions via unbalanced optimal transport and measures their causal effect on model outputs through targeted interventions. MedFocus produces spatial, concept-level, and token-level attributions and substantially outperforms prior methods, taking a step toward more trustworthy attribution for medical LVLMs. Our data and code are available at https://github.com/gzxiong/medfocus/.

Vision-language alignment using chest X-rays and radiology reports has emerged as an advanced paradigm for zero-shot classification and grounding of chest X-ray findings. However, standard contrastive learning typically treats radiographs and reports from different patients simply as negative pairs. This assumption introduces noisy negatives, as different patients frequently exhibit similar findings. Such noisy negatives cause semantic ambiguity and degrade performance in zero-shot understanding tasks. To address this challenge, we propose CoNNS, a concept-guided noisy-negative suppression framework. To support the negative suppression mechanism, unlike previous methods that use raw reports or templatized texts, we construct a hierarchical concept ontology using large language models. The ontology structures 41 key clinical concepts by explicitly modeling presence, attributes (location and characteristics), and texts (evidential segment and presence statement). Leveraging this ontology, we implement a cross-patient pair relabeling strategy comprising three steps: (1) Fine-Grained Breakdown to categorize pairs based on finding presence; (2) Noisy Negative Filtering to resolve semantic conflicts by removing false negatives; and (3) Hard Negative Mining to identify subtle attribute discrepancies using a lightweight language model. Finally, we propose a Concept-Aware NCE loss to align visual features with text while suppressing the identified noisy negatives. Extensive experiments across multi-granularity zero-shot grounding tasks and five zero-shot classification datasets validate that CoNNS outperforms existing state-of-the-art models. The code is available at https://github.com/DopamineLcy/conns.

Deep learning methods have demonstrated promising results in predicting BI-RADS scores from mammography images. However, the interpretation of these images can vary, leading to discrepancies even among radiologists. Given the inherent complexity of mammograms, training classification models solely on image labels often yields limited performance. To address this challenge, we curated 2313 mammogram images and their corresponding captions from two mammography atlases. Our proposed approach employs a multi-modal model that uses a pretrained PubMedBERT as the language component. By training this model on image-text pairs with contrastive learning, we enable the vision encoder to absorb the rich information contained in the captions, thereby improving its understanding of mammography findings. We then fine-tune the vision encoder on two datasets for BI-RADS prediction, achieving superior performance compared with models trained without this pretraining, particularly when labeled samples are scarce. The improvement in the 3-class average F1 score ranges from +1% to +14%: a +1% increase with 40K training samples, and a +14% increase with 1K samples. Furthermore, our experiments reveal that 2K image-text pairs from mammography atlases can be more informative than 2K labeled samples for label prediction, with an average margin of +1.1% when more than 10K training samples are available. Overall, our work provides a vision-language model for mammography and highlights the value of textual information from mammography atlases. In addition, we publicly release preprocessed mammography images of the TEKNOFEST dataset. The training code, pre-trained model weights, data extraction scripts, and the released dataset are publicly available at: https://github.com/igulluk/MAM-CLIP

Medical foundation models have achieved remarkable clinical performance, yet their robustness under real-world perturbations remains underexplored. We present a robustness benchmark comprising 40 perturbation types (12 base, 28 medical-specific) across eight imaging modalities, evaluating five VLMs (LLaVA-Med, MedGemma, MedGemma-1.5, Gemini-2.5-flash and GPT-4o-mini) on VQA, visual grounding, and captioning, alongside two segmentation models (MedSAM, SAM-Med2D) with five fine-tuning strategies. Our findings reveal: (1) Fine-tuning strategy dominates robustness, with LoRA exhibiting nearly double the degradation of full fine-tuning, while SAM-Med2D's Adapter offers favorable efficiency-robustness trade-off. (2) Medical-specific perturbations disproportionately damage segmentation, with 9 of 15 top corruptions being domain-specific. (3) LoRA-tuned visual grounding drops over 40 points, whereas zero-shot captioning remains stable (<7% drop). Zero-shot VQA shows model-dependent robustness--medical models drop under 20% while Gemini-2.5-flash drops 54%. General-purpose VLMs achieve higher VQA accuracy but fail on grounding; among medical VLMs, MedGemma demonstrates the best overall stability. These results provide deployment guidelines and underscore the necessity of domain-specific robustness evaluation for medical AI. Our code is available at: https://abnerai.github.io/MedFM-Robust.

Vision-language models (VLMs) can couple visual perception with open-ended clinical reasoning, making them attractive for computational histopathology. However, fine-tuning billions of parameters on scarce, expert-annotated pathology data is prohibitive, while in-context learning (ICL), which conditions the VLM on demonstrative image-text pairs without parameter updates, suffers from high sensitivity to which examples are selected and how the query is phrased, producing unreliable diagnostics. Existing selection strategies rely on query-dependent nearest-neighbour retrieval that ignores global data structure, require costly parameter updates, or disregard the joint vision-text embedding geometry of VLMs. We propose GAUC, a training-free coreset selection method operating directly in the pre-trained multimodal embedding space. GAUC jointly optimises three objectives: (1) a Maximum Mean Discrepancy term enforcing distributional fidelity between coreset and full dataset, (2) an Effective Mutual Information Difference regulariser bounding performance degradation under prompt paraphrases by exploiting the VLM's joint vision-text alignment, and (3) a predictive-uncertainty (entropy) penalty suppressing ambivalent, hallucination-prone outputs. On CRC-100K and MHIST across multiple open-source VLM architectures, GAUC \emph{matches} the accuracy of the strongest ICL selection and dataset-distillation baselines while substantially improving calibration, prompt robustness, and hallucination rates, all without a single gradient update.

Small vision-language models (2-8B) are well-suited for clinical deployment due to privacy constraints, limited connectivity, and low-latency requirements favouring on-device or on-premise inference. However, their limited capacity exacerbates the generation of plausible but incorrect outputs. We extend game-theoretic decoding, previously restricted to text-only, closed-ended NLP tasks, to vision-language models for open-ended Medical VQA. We introduce a semantically aware Wasserstein stopping criterion that replaces lexical order matching, enabling convergence based on semantic consensus among near-synonymous candidate answers and avoiding unnecessary iterations caused by clinically equivalent ranking swaps. On VQA-RAD and PathVQA, we obtain consistent, statistically significant improvements over greedy and discriminative baselines. On VQA-RAD, we improve Qwen3-VL-2B by +3.5 percentage points (p < 0.01), surpassing the greedy 4B model, with similar trends at larger scales. On PathVQA, Gemma-3-4B with BDG matches MedGemma-4B under greedy decoding despite no domain-specific fine-tuning. At accuracy parity with classic BDG, the Wasserstein criterion reduces average convergence iterations by approximately 20%, improving inference efficiency while preserving the game-theoretic equilibrium behaviour. Code is available at https://github.com/luca-hagen/ Wasserstein-BDG-medical-VQA.

Recent advancements in Large Language Models (LLMs) and Large Vision Language Models (LVLMs) have enabled general-purpose systems to demonstrate promising capabilities in complex reasoning tasks, including those in the medical domain. Medical Visual Question Answering (MedVQA) has particularly benefited from these developments. However, despite Bangla being one of the most widely spoken languages globally, there exists no established MedVQA benchmark for it. To address this gap, we introduce BanglaMedVQA, a dataset comprising clinically validated image-question-answer pairs, along with a comprehensive evaluation of current foundation models on this resource. Consistent with prior findings that report low performance of current models on English MedVQA benchmarks, our analysis reveals that Bangla performance is substantially lower, reflecting the challenges inherent to low-resource languages. Even top-performing models such as Gemini and GPT-4.1 mini fail to accurately answer specialized diagnostic questions, indicating severe limitations in fine-grained medical reasoning. Although certain open-source models, such as Gemma-3, occasionally outperform these models in general categories, they too struggle with clinically complex questions, underscoring the urgent need for top-notch evaluation method.

Recent progress in deep learning has significantly advanced CT image analysis, particularly for segmentation tasks. However, these advances are largely confined to image-level pattern recognition, with most methods lacking explicit anatomical or contextual reasoning. Large vision-language models introduce linguistic context into image analysis, yet most approaches typically focus on a single task, which is insufficient for clinical workflow analysis that requires multiple fine-grained types of analysis, such as anatomy detection and segmentation. In this paper, we propose a unified autoregressive framework that integrates language-guided visual reasoning into CT interpretation. Our method introduces task-routing tokens that trigger detection and segmentation heads conditioned on the hidden states of a large vision-language model, enabling coherent generation of visual outputs (e.g., masks and bounding boxes) and textual reasonings. To progressively enhance localisation accuracy and semantic clarity, we further design a "closer-look" mechanism that allows the model to perform progressive coarse-to-fine visits to regions of interest under refined fields of view. To support model training and evaluation, we curated a new multimodal CT dataset containing pixel-wise masks, bounding boxes, spatial prompts, and structured descriptions for visual objects constructed through an AI-assisted annotation process with human verification. Experiments on public benchmarks demonstrate consistent improvements over the SoTA, achieving up to 1.0% Dice on BTCV and 1.7% Dice on MosMed+, while additionally providing appearance reasoning outputs. The code and dataset will be available.

Biomedical Vision--Language Models (VLMs) have shown remarkable promise in few-shot medical diagnosis but face a critical bottleneck: \textit{fragility to prompt variations}.Existing adaptation frameworks typically optimize visual and textual prompts as independent streams, relying on ideal ``Golden Prompts''. In clinical reality, where descriptions are often noisy and heterogeneous, this modality isolation leads to unstable cross-modal alignment. To address this, we propose BiomedAP, a vision-informed dual-anchor framework with gated cross-modal fusion.BiomedAP enforces synergistic alignment through two mechanisms: (1) Gated Cross-Modal Fusion, which enables layer-wise interaction between modalities, acting as a dynamic noise regulator to suppress irrelevant textual cues; and (2) a Dual-Anchor Constraint that regularizes learnable prompts toward stable semantic centroids derived from both expert templates (High Anchors) and few-shot visual prototypes (Low Anchors). Extensive experiments across 11 benchmarks demonstrate that BiomedAP consistently surpasses baselines, achieving competitive few-shot accuracy and markedly enhanced robustness under prompt perturbations. Our code is available at: https://github.com/tongdiedie/BiomedAP. Keywords: Vision-Language Models; Prompt Learning; Parameter-Efficient Fine-Tuning; Few-shot Learning

Vision-Language Models (VLMs) facilitate medical visual question answering (MedVQA) by jointly interpreting images and text. However, existing models typically depend on large architectures and closed-set answers, which limits their efficiency and potential clinical applicability. To overcome these shortcomings, we introduce RoiMAM, an efficient VLM. It integrates a training-free ROI Generation Module with Semantic Selective Suppression to focus on lesion-relevant regions, alongside a Text Prompt Enhancer module that provides modality-specific context without introducing training parameters. Compared to the widely used MedVInT-TD model, our design achieves efficient and accurate diagnosis at less than 20% of the model size, while improving accuracy by approximately 2% on SLAKE and 4.6% on PMC-VQA.

Dermatological diagnosis requires integrating fine-grained visual perception with expert clinical knowledge. Although Multimodal Large Language Models (MLLMs) facilitate interactive medical image analysis, their application in dermatology is hindered by insufficient domain-specific grounding and hallucinations. To address these issues, we propose DermAgent, a collaborative multi-tool agent that orchestrates seven specialized vision and language modules within a Plan-Execute-Reflect framework. DermAgent delivers stepwise, traceable diagnostic reasoning through three core components. First, it employs complementary visual perception tools for comprehensive morphological description, dermoscopic concept annotation, and disease diagnosis. Second, to overcome the lack of domain prior, a dual-modality retrieval module anchors every prediction in external evidence by cross-referencing 413,210 diagnosed image cases and 3,199 clinical guideline chunks. To further mitigate hallucinations, a deterministic critic module conducts strict post-hoc auditing via confidence, coverage, and conflict gates, automatically detecting inter-source disagreements to trigger targeted self-correction. Extensive experiments on five dermatology benchmarks demonstrate that DermAgent consistently outperforms state-of-the-art MLLMs and medical agent baselines across zero-shot fine-grained disease diagnosis, concept annotation, and clinical captioning tasks, exceeding GPT-4o by 17.6% in skin disease diagnostic accuracy and 3.15% in captioning ROUGE-L. Our code is available at https://github.com/YizeezLiu/DermAgent.

Fine-grained Vision-Language Pre-training (FVLP) demonstrates significant potential in 3D medical image understanding by aligning anatomy-level visual representations with corresponding textual descriptions. However, existing FVLP paradigms often suffer from severe representation collapse in the textual embedding space, where text embeddings of distinct anatomical structures become highly clustered and indistinguishable. This distributional degeneracy renders the model hypersensitive to prompt variations, hindering reliable clinical deployment. To address these challenges, we propose a novel Cross-Anatomy Global-Local Contrastive Learning framework (CA-GCL). CA-GCL introduces a global contrastive objective that enforces separation between anatomical categories in the latent space, effectively counteracting the aggregation tendency induced by local alignment. Furthermore, we incorporate a clinical-aware text augmentation strategy based on permutation invariance and partial completeness to enhance robustness against descriptive incompleteness. Extensive evaluations on the CT-RATE and Rad-ChestCT datasets show that CA-GCL achieves comparable zero-shot abnormality detection performance to existing VLP paradigms, while demonstrating substantially better robustness to prompt variations: on canonical templates it obtains higher mean AUC with lower variance, and on non-canonical templates it remains stable whereas baselines degrade markedly. These results validate CA-GCL as an effective framework for robust 3D medical image understanding.

Self-verification, re-invoking the same vision language model (VLM) in a fresh context to check its own generated answer, is increasingly used as a default safety layer for medical visual question answering (VQA). We argue that this practice is fundamentally unreliable. We introduce [METHOD NAME], a diagnostic framework for mapping the reliability boundary of medical VLM self-verification by decomposing verifier behavior into discrimination capability and agreement bias. Because the verifier and answer generator are capacity-coupled, the verifier can overly agree with the generator, creating a verification mirage: a regime with both high verifier error and high agreement bias, driven by false acceptance of incorrect answers. Evaluating six open-weight VLMs across five medical VQA datasets and seven medical tasks, we find that this boundary is strongly task-conditioned. Knowledge-intensive clinical tasks fall deepest into the mirage, simpler tasks are more resistant, and perceptual tasks lie in between. Verification also fails to provide an independent safety signal: logistic mixed-effects analysis shows that verifier error and agreement bias become more likely when the generator is wrong, while saliency analyses show that verifiers under-attend to image evidence relative to generators, a phenomenon we call the lazy verifier. Cross-verification reduces but does not eliminate the mirage. Moreover, when verification is reused in multi-turn actor-verifier loops, most initially wrong answers become locked in by false verification. Since our experiments use clean benchmarks, the observed reliability boundary likely underestimates failures in real clinical deployment.

Large vision-language models (VLMs) demonstrate strong performance in medical image understanding, but frequently generate clinically plausible yet incorrect statements, raising significant safety concerns. Existing medical hallucination benchmarks primarily focus on 2D imaging with one-shot diagnostic questions, offering limited insight into whether predictions are grounded in correct localization and abnormality identification, allowing critical reasoning errors to remain hidden behind seemingly correct diagnoses. We introduce Med-StepBench, the first large-scale benchmark for step-wise hallucination detection in 3D oncological PET/CT, comprising over 12,000 images and more than 1,000,000 image-statement pairs across volumetric and multi-view 2D data, which decomposes clinical reasoning into four expert-designed diagnostic stages. Using clinician-verified annotations, we perform the first step-level evaluation of general-purpose and medical VLMs, revealing systematic failure modes obscured by aggregate accuracy metrics. Furthermore, we show that current VLMs are highly susceptible to adversarial yet clinically plausible intermediate explanations, which significantly amplify hallucinations despite contradictory visual evidence. Together, our findings highlight fundamental limitations in grounding multi-step clinical reasoning and establish Med-StepBench as a rigorous benchmark for developing safer and more reliable medical VLMs.

The reasoning gap between large and compact vision-language models (VLMs) limits the deployment of medical AI on portable clinical devices. Compact VLMs of 2--4B parameters can run on resource-constrained hardware but lack the multi-step reasoning capacity needed for interpretable clinical decision support. Existing knowledge distillation methods transfer answers without the reasoning process behind them. Medical visual question answering (VQA) serves as a testbed for this problem, as it requires models to integrate visual evidence with clinical knowledge through structured reasoning chains. We introduce LiteMedCoT-VL, a pipeline that transfers chain-of-thought reasoning from a 235B teacher model to 2B student models through LoRA-based fine-tuning on explanation-enriched training data. All inference is conducted without image captions by default, simulating the clinical scenario in which a physician interprets a medical image directly without an accompanying radiology report. On the PMC-VQA benchmark, LiteMedCoT-VL achieves 64.9% accuracy, exceeding the zero-shot Qwen3-VL-4B baseline of 53.9% by 11.0 percentage points and outperforming all published baselines. This result indicates that a 2B model with reasoning distillation can match or exceed models with twice the parameters. Visual grounding analysis shows that the model relies on image content rather than exploiting textual priors. Our code is publicly available at https://anonymous.4open.science/r/LiteMedCoT-VL.

Automated grading of diabetic retinopathy (DR) faces several critical challenges: subtle inter-grade visual distinctions in fine-grained lesion patterns, distributional discrepancies induced by heterogeneous imaging devices and acquisition conditions, and the inherent inability of purely visual approaches to exploit clinical semantic knowledge. In this paper, we propose CLIP-Guided Semantic Diffusion (CGSD), a DR grading framework that synergistically integrates vision-language pretraining with diffusion probabilistic modeling. We adopt a domain-specific vision-language model tailored for DR grading as the semantic guidance module and adapt it to the target domain via Low-Rank Adaptation (LoRA), effectively bridging the distributional gap between the pretrained model and the target dataset with only a minimal number of trainable parameters. Building on this foundation, we construct a cross-modal semantic conditioning vector by computing the dot product between image features and the text description features of each DR grade, yielding a joint representation that simultaneously encodes visual content and clinical-grade semantics. This vector serves as the conditioning signal for the diffusion denoising network, replacing the structurally complex dual-branch visual prior employed in existing diffusion-based classification methods. Experiments on the APTOS 2019 dataset demonstrate that the proposed approach achieves an accuracy of 87.5% and a macro-averaged F1 score of 0.731, outperforming a variety of representative methods. Ablation studies further validate the independent contribution of each constituent module.

Wireless capsule endoscopy (WCE) enables non-invasive visual assessment of the small bowel, but its clinical utility is constrained by the large volume of frames generated per examination and the difficulty of recognising subtle abnormalities under highly variable imaging conditions. Existing learning-based approaches for WCE are predominantly vision-only, often confined to narrow pathology sets, and show limited transfer across datasets and centres. To address these limitations, this study introduces CapCLIP, a domain-specific vision-language representation learning framework for WCE. CapCLIP aligns capsule endoscopy frames with clinically grounded textual descriptions derived from standardised nomenclature and pathology-aware caption templates, thereby learning embeddings that are both semantically informed and transferable. The proposed framework is evaluated against relevant open-source vision and vision-language foundation models under strict zero-shot conditions using unseen WCE datasets. Evaluation covers three downstream tasks: K-nearest neighbour classification, CLIP-style image-text classification, and text-to-image retrieval. Across these settings, CapCLIP consistently outperforms the compared baselines, with particularly strong gains in zero-shot image-text classification and cross-modal retrieval on out-of-distribution datasets. The results indicate that language-guided representation learning can improve both generalisation and semantic interpretability in WCE analysis. These findings position CapCLIP as a step toward foundation models tailored to capsule endoscopy and support the use of language-grounded WCE analysis.

Medical vision--language models (VLMs) are usually evaluated on intact image--question pairs, but trustworthy clinical use requires a stronger property: a model must recognise when the evidential basis for an answer has failed. We study this through silent failures under perturbed evidence, where a vision-required medical question is paired with a false premise, wording perturbation, knowledge-only rewrite, or ROI-corrupted image, yet the model returns a fluent non-refusal answer. We introduce medvigil, a 300-case evaluation suite drawn from four public medical VQA sources, supervised end to end by four board-certified radiologists: every gold answer, refusal option, candidate-answer set, paraphrase, false-premise trap, ROI box, and clinical risk tier is clinician-authored. Two attending radiologists annotate every case in parallel, a senior radiologist consolidates the released manifest, and a separate fourth radiologist independent of construction answers every probe to provide the human reference baseline. The release contains 2556 MCQ probes, 240 counterfactual triplets, physician-adjudicated risk-tier and answerability flags, ROI boxes, and a paired open-ended variant. We report seven correctness-conditioned audit metrics that summarise into the medvigil Composite Score (MCS), and audit 16 vision-capable models plus two text-only baselines. The independent radiologist scores MCS 83.3 at silent-failure rate 5.8%, leaving a 14.1-point composite headroom above the strongest audited model (Claude Opus 4.7 at 69.2). The benchmark and evaluation harness are publicly released.

Diabetic Retinopathy (DR) is a leading cause of preventable blindness among working-age adults worldwide, yet most automated screening systems are limited to image-level classification and lack clinically structured reporting. We propose Retina-RAG, a low-cost modular framework that jointly performs DR severity grading, macular edema (ME) detection, and report generation. The architecture decouples a high-performance retinal classifier and a parameter-efficient vision-language model (Qwen2.5-VL-7B-Instruct) adapted via Low-Rank Adaptation (LoRA), enabling flexible component integration. A retrieval-augmented generation (RAG) module injects curated ophthalmic knowledge together with structured classifier outputs at inference time to improve diagnostic consistency and reduce hallucinations. Retina-RAG achieves an F1-score of 0.731 for DR grading and 0.948 for ME detection, substantially outperforming zero-shot Qwen (0.096, 0.732) and MMed-RAG (0.541, 0.641) on a retinal disease detection dataset with captions. For report generation, Retina-RAG attains ROUGE-L 0.438 and SBERT similarity 0.884, exceeding all baselines. The full framework operates on a single consumer-grade GPU, demonstrating that clinically structured retinal AI can be achieved with modest computational resources.

Automated glaucoma detection is critical for preventing irreversible vision loss and reducing the burden on healthcare systems. However, ensuring fairness across diverse patient populations remains a significant challenge. In this paper, we propose FairEnc, a fair pretraining method for vision-language models (VLMs) that enables simultaneous debiasing across multiple sensitive attributes. FairEnc jointly mitigates biases in both textual and visual modalities with respect to multiple sensitive attributes, including race, gender, ethnicity, and language. Specifically, for the textual encoder, we leverage a large language model to generate synthetic clinical descriptions with varied sensitive attributes while preserving disease semantics, and employ a contrastive alignment objective to encourage demographic-invariant representations. For the visual encoder, we propose a dual-level fairness strategy that combines mutual information regularization to reduce statistical dependence between learned features and demographic groups, with multi-discriminator adversarial debiasing. Comprehensive experiments on the publicly available Harvard-FairVLMed dataset demonstrate that FairEnc effectively reduces demographic disparity as measured by DPD and DEOdds while achieving strong diagnostic performance under both zero-shot and linear probing evaluations. Additional experiments on the private FairFundus dataset show that FairEnc consistently preserves fairness advantages under cross-domain and cross-modality settings and maintains diagnostic performance within a competitive range. These results highlight FairEnc's ability to generalize fairness under distribution shifts, supporting its potential for more equitable deployment in real-world clinical settings. Our codebase and synthetic clinical notes are available at https://github.com/Mohamed-Elhabebe/FairEnc

Access to diverse, well-annotated medical images with interactive learning tools is fundamental for training practitioners in medicine and related fields to improve their diagnostic skills and understanding of anatomical structures. While medical atlases are valuable, they are often impractical due to their size and lack of interactivity, whereas online image search may provide mislabeled or incomplete material. To address this, we propose MIRAGE, a multimodal medical text and image retrieval and generation system that allows users to find and generate clinically relevant images from trustworthy sources by mapping both text and images to a shared latent space, enabling semantically meaningful queries. The system is based on a fine-tuned medical version of CLIP (MedICaT-ROCO), trained with the ROCO dataset, obtained from PubMed Central. MIRAGE allows users to give prompts to retrieve images, generate synthetic ones through a medical diffusion model (Prompt2MedImage) and receive enriched descriptions from a large language model (Dolly-v2-3b). It also supports a dual search option, enabling the visual comparison of different medical conditions. A key advantage of the system is that it relies entirely on publicly available pretrained models, ensuring reproducibility and accessibility. Our goal is to provide a free, transparent and easy-to-use didactic tool for medical students, especially those without programming skills. The system features an interface that enables interactive and personalized visual learning through medical image retrieval and generation. The system is accessible to medical students worldwide without requiring local computational resources or technical expertise, and is currently deployed on Kaggle: http://www-vpu.eps.uam.es/mirage