Myeongkyun Kang

Gexin Huang, Yanting Yang, Myeongkyun Kang et al.

Vision-language models (VLMs) excel on visual question answering and multimodal reasoning benchmarks. Yet their capability on ultra-resolution images - where critical evidence is tiny, subtle, spatially distant, or distributed - remains unclear. Existing evaluations largely report final-answer accuracy, offering limited insight into whether models acquire and integrate the necessary visual evidence. We introduce UltraVR, a diagnostic benchmark for evidence-grounded visual reasoning over ultra-resolution images. UltraVR spans four high-value scenarios: CCTV surveillance, remote sensing (RS), whole-slide image (WSI) pathology, and industrial anomaly detection (AD). These domains pose complementary challenges: fine-grained object grounding in crowded CCTV scenes, long-range spatial comparison in RS, multi-scale evidence navigation in WSI, and subtle irregularity detection in repetitive industrial layouts. Beyond standard QA triples, each instance includes a structured ground-truth chain of thought with step-level questions, intermediate answers, and reasoning labels. These labels decompose reasoning into evidence grounding, local perception, quantification, evidence integration, and decision inference, enabling process-level diagnosis over black-box scoring. Using UltraVR, we evaluate frontier VLMs and show that current models remain far from reliable on ultra-resolution reasoning. Importantly, the structured annotations allow us to localize failures across the visual-to-decision pipeline: errors concentrate in evidence grounding and local perception, while downstream inference often recovers when intermediate visual facts are supplied. These findings demonstrate UltraVR as a diagnostic testbed for measuring not only whether VLMs answer correctly, but where their ultra-resolution reasoning process breaks.

Myeongkyun Kang, Soopil Kim, Xiaoxiao Li et al.

Large vision language models (LVLMs) have demonstrated impressive performance across a wide range of tasks. These capabilities largely stem from visual instruction tuning, which fine-tunes models on datasets consisting of curated image-instruction-output triplets. However, in the medical domain, constructing large-scale, high-quality instruction datasets is particularly challenging due to the need for specialized expert knowledge. To address this issue, we propose an instruction-free tuning approach that reduces reliance on handcrafted instructions, leveraging only image-description pairs for fine-tuning. Specifically, we introduce a momentum proxy instruction as a replacement for curated text instructions, which preserves the instruction-following capability of the pre-trained LVLM while promoting updates to parameters that remain valid during inference. Consequently, the fine-tuned LVLM can flexibly respond to domain-specific instructions, even though explicit instructions are absent during fine-tuning. Additionally, we incorporate a response shuffling strategy to mitigate the model's over-reliance on previous words, facilitating more effective fine-tuning. Our approach achieves state-of-the-art accuracy on multiple-choice visual question answering tasks across SKINCON, WBCAtt, CBIS, and MIMIC-CXR datasets, significantly enhancing the fine-tuning efficiency of LVLMs in medical domains.

Myeongkyun Kang, Yanting Yang, Xiaoxiao Li

Fine-grained representation learning is crucial for retrieval and phrase grounding in chest X-rays, where clinically relevant findings are often spatially confined. However, the lack of region-level supervision in contrastive models and the limited ability of large vision language models to capture fine-grained representations in external validation lead to suboptimal performance on these tasks. To address these limitations, we propose Location-aware Fine-grained representation learning (LoFi), which jointly optimizes sigmoid, captioning, and location-aware captioning losses using a lightweight large language model. The location-aware captioning loss enables region-level supervision through grounding and dense captioning objectives, thereby facilitating fine-grained representation learning. Building upon these representations, we integrate a fine-grained encoder into retrieval-based in-context learning to enhance chest X-ray grounding across diverse settings. Extensive experiments demonstrate that our method achieves superior retrieval and phrase grounding performance on MIMIC-CXR and PadChest-GR.

Sion An, Myeongkyun Kang, Soopil Kim et al.

Electroencephalography (EEG) motor imagery (MI) classification is a fundamental, yet challenging task due to the variation of signals between individuals i.e., inter-subject variability. Previous approaches try to mitigate this using task-specific (TS) EEG signals from the target subject in training. However, recording TS EEG signals requires time and limits its applicability in various fields. In contrast, resting state (RS) EEG signals are a viable alternative due to ease of acquisition with rich subject information. In this paper, we propose a novel subject-adaptive transfer learning strategy that utilizes RS EEG signals to adapt models on unseen subject data. Specifically, we disentangle extracted features into task- and subject-dependent features and use them to calibrate RS EEG signals for obtaining task information while preserving subject characteristics. The calibrated signals are then used to adapt the model to the target subject, enabling the model to simulate processing TS EEG signals of the target subject. The proposed method achieves state-of-the-art accuracy on three public benchmarks, demonstrating the effectiveness of our method in cross-subject EEG MI classification. Our findings highlight the potential of leveraging RS EEG signals to advance practical brain-computer interface systems. The code is available at https://github.com/SionAn/MICCAI2024-ResTL.

Ruinan Jin, Beidi Zhao, Myeongkyun Kang et al.

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.

Soopil Kim, Sion An, Philip Chikontwe et al.

Logical anomalies (LA) refer to data violating underlying logical constraints e.g., the quantity, arrangement, or composition of components within an image. Detecting accurately such anomalies requires models to reason about various component types through segmentation. However, curation of pixel-level annotations for semantic segmentation is both time-consuming and expensive. Although there are some prior few-shot or unsupervised co-part segmentation algorithms, they often fail on images with industrial object. These images have components with similar textures and shapes, and a precise differentiation proves challenging. In this study, we introduce a novel component segmentation model for LA detection that leverages a few labeled samples and unlabeled images sharing logical constraints. To ensure consistent segmentation across unlabeled images, we employ a histogram matching loss in conjunction with an entropy loss. As segmentation predictions play a crucial role, we propose to enhance both local and global sample validity detection by capturing key aspects from visual semantics via three memory banks: class histograms, component composition embeddings and patch-level representations. For effective LA detection, we propose an adaptive scaling strategy to standardize anomaly scores from different memory banks in inference. Extensive experiments on the public benchmark MVTec LOCO AD reveal our method achieves 98.1% AUROC in LA detection vs. 89.6% from competing methods.

Myeongkyun Kang, Dongkyu Won, Miguel Luna et al.

Models trained on datasets with texture bias usually perform poorly on out-of-distribution samples since biased representations are embedded into the model. Recently, various image translation and debiasing methods have attempted to disentangle texture biased representations for downstream tasks, but accurately discarding biased features without altering other relevant information is still challenging. In this paper, we propose a novel framework that leverages image translation to generate additional training images using the content of a source image and the texture of a target image with a different bias property to explicitly mitigate texture bias when training a model on a target task. Our model ensures texture similarity between the target and generated images via a texture co-occurrence loss while preserving content details from source images with a spatial self-similarity loss. Both the generated and original training images are combined to train improved classification or segmentation models robust to inconsistent texture bias. Evaluation on five classification- and two segmentation-datasets with known texture biases demonstrates the utility of our method, and reports significant improvements over recent state-of-the-art methods in all cases.

Myeongkyun Kang, Philip Chikontwe, Miguel Luna et al.

Following the pandemic outbreak, several works have proposed to diagnose COVID-19 with deep learning in computed tomography (CT); reporting performance on-par with experts. However, models trained/tested on the same in-distribution data may rely on the inherent data biases for successful prediction, failing to generalize on out-of-distribution samples or CT with different scanning protocols. Early attempts have partly addressed bias-mitigation and generalization through augmentation or re-sampling, but are still limited by collection costs and the difficulty of quantifying bias in medical images. In this work, we propose Mixing-AdaSIN; a bias mitigation method that uses a generative model to generate de-biased images by mixing texture information between different labeled CT scans with semantically similar features. Here, we use Adaptive Structural Instance Normalization (AdaSIN) to enhance de-biasing generation quality and guarantee structural consistency. Following, a classifier trained with the generated images learns to correctly predict the label without bias and generalizes better. To demonstrate the efficacy of our method, we construct a biased COVID-19 vs. bacterial pneumonia dataset based on CT protocols and compare with existing state-of-the-art de-biasing methods. Our experiments show that classifiers trained with de-biased generated images report improved in-distribution performance and generalization on an external COVID-19 dataset.