arXiv subjects

Sungmin Hong

Publications and source records attributed to Sungmin Hong.

5 recordsLinked to original sources

arXiv2026

Quantifying Consistency in LLM Logical Reasoning via Structural Uncertainty

Large language models can arrive at the same answer through reasoning paths that are unstable, contradictory, or difficult to rank consistently -- a failure mode especially prevalent in multi-step deductive reasoning. Existing methods assess reliability primarily through output dispersion -- measuring how much sampled answers differ -- but this discards a complementary signal: whether the model can consistently rank competing reasoning candidates. We propose structural uncertainty, a consistency-aware framework derived from the stability of self-preference-induced rankings over sampled reasoning solutions. Given a query, we generate multiple candidate solutions and ask the model to judge pairwise preferences among its own outputs. We aggregate self-preferences into ranking distributions via Bradley-Terry modeling with PageRank, and decompose the signal into two entropy-based components: across-trial ranking instability and within-trial candidate ambiguity. Across five LLMs and eight benchmarks, structural signals provide information complementary to answer dispersion: on logical and mathematical reasoning tasks, the combination improves identification of unreliable instances, while on factual retrieval the structural signal collapses toward uniformity, diagnosing a regime boundary where reasoning-level consistency evaluation is uninformative. The two components relate differently to accuracy: within-trial ambiguity correlates positively with correctness -- consistent with settings where multiple plausible solution paths remain competitive -- while across-trial instability correlates negatively, signaling unreliable reasoning. Structural uncertainty is best understood not as a universal confidence estimator, but as a regime-sensitive evaluator of logical reasoning consistency.

cs.AI↗

arXiv2025

CompAgent: An Agentic Framework for Visual Compliance Verification

Visual compliance verification is a critical yet underexplored problem in computer vision, especially in domains such as media, entertainment, and advertising where content must adhere to complex and evolving policy rules. Existing methods often rely on task-specific deep learning models trained on manually labeled datasets, which are costly to build and limited in generalizability. While recent Multimodal Large Language Models (MLLMs) offer broad real-world knowledge and policy understanding, they struggle to reason over fine-grained visual details and apply structured compliance rules effectively on their own. In this paper, we propose CompAgent, the first agentic framework for visual compliance verification. CompAgent augments MLLMs with a suite of visual tools-such as object detectors, face analyzers, NSFW detectors, and captioning models-and introduces a planning agent that dynamically selects appropriate tools based on the compliance policy. A compliance verification agent then integrates image, tool outputs, and policy context to perform multimodal reasoning. Experiments on public benchmarks show that CompAgent outperforms specialized classifiers, direct MLLM prompting, and curated routing baselines, achieving up to 76% F1 score and a 10% improvement over the state-of-the-art on the UnsafeBench dataset. Our results demonstrate the effectiveness of agentic planning and robust tool-augmented reasoning for scalable, accurate, and adaptable visual compliance verification.

cs.CV↗

arXiv2023

Effectively Fine-tune to Improve Large Multimodal Models for Radiology Report Generation

Writing radiology reports from medical images requires a high level of domain expertise. It is time-consuming even for trained radiologists and can be error-prone for inexperienced radiologists. It would be appealing to automate this task by leveraging generative AI, which has shown drastic progress in vision and language understanding. In particular, Large Language Models (LLM) have demonstrated impressive capabilities recently and continued to set new state-of-the-art performance on almost all natural language tasks. While many have proposed architectures to combine vision models with LLMs for multimodal tasks, few have explored practical fine-tuning strategies. In this work, we proposed a simple yet effective two-stage fine-tuning protocol to align visual features to LLM's text embedding space as soft visual prompts. Our framework with OpenLLaMA-7B achieved state-of-the-art level performance without domain-specific pretraining. Moreover, we provide detailed analyses of soft visual prompts and attention mechanisms, shedding light on future research directions.

cs.CV↗

arXiv2021

Hypernet-Ensemble Learning of Segmentation Probability for Medical Image Segmentation with Ambiguous Labels

Despite the superior performance of Deep Learning (DL) on numerous segmentation tasks, the DL-based approaches are notoriously overconfident about their prediction with highly polarized label probability. This is often not desirable for many applications with the inherent label ambiguity even in human annotations. This challenge has been addressed by leveraging multiple annotations per image and the segmentation uncertainty. However, multiple per-image annotations are often not available in a real-world application and the uncertainty does not provide full control on segmentation results to users. In this paper, we propose novel methods to improve the segmentation probability estimation without sacrificing performance in a real-world scenario that we have only one ambiguous annotation per image. We marginalize the estimated segmentation probability maps of networks that are encouraged to under-/over-segment with the varying Tversky loss without penalizing balanced segmentation. Moreover, we propose a unified hypernetwork ensemble method to alleviate the computational burden of training multiple networks. Our approaches successfully estimated the segmentation probability maps that reflected the underlying structures and provided the intuitive control on segmentation for the challenging 3D medical image segmentation. Although the main focus of our proposed methods is not to improve the binary segmentation performance, our approaches marginally outperformed the state-of-the-arts. The codes are available at \url{https://github.com/sh4174/HypernetEnsemble}.

eess.IV↗

arXiv2021

3D-StyleGAN: A Style-Based Generative Adversarial Network for Generative Modeling of Three-Dimensional Medical Images

Image synthesis via Generative Adversarial Networks (GANs) of three-dimensional (3D) medical images has great potential that can be extended to many medical applications, such as, image enhancement and disease progression modeling. However, current GAN technologies for 3D medical image synthesis need to be significantly improved to be readily adapted to real-world medical problems. In this paper, we extend the state-of-the-art StyleGAN2 model, which natively works with two-dimensional images, to enable 3D image synthesis. In addition to the image synthesis, we investigate the controllability and interpretability of the 3D-StyleGAN via style vectors inherited form the original StyleGAN2 that are highly suitable for medical applications: (i) the latent space projection and reconstruction of unseen real images, and (ii) style mixing. We demonstrate the 3D-StyleGAN's performance and feasibility with ~12,000 three-dimensional full brain MR T1 images, although it can be applied to any 3D volumetric images. Furthermore, we explore different configurations of hyperparameters to investigate potential improvement of the image synthesis with larger networks. The codes and pre-trained networks are available online: https://github.com/sh4174/3DStyleGAN.

eess.IV↗