arXiv subjects

Yugyung Lee

Publications and source records attributed to Yugyung Lee.

At least 19 recordsLinked to original sources

arXiv2026

SEMIR: Topology-Preserving Graph Minors for Thin-Structure Segmentation

Thin-structure segmentation--power lines, cracks, lane markings at 1-3 pixel width--requires preserving connectivity that standard representations preclude: patching severs continuous structures and conventional superpixels merge thin targets into background before classification. Topology-aware losses penalize connectivity breaks at the objective level but cannot recover what the representation has already destroyed. We propose SEMIR, a framework that replaces the pixel lattice with a parameterized graph minor whose contraction map preserves thin-structure connectivity under the contraction criterion. The minor collapses millions of pixels into tens or hundreds of boundary-aligned supernodes, enabling full-resolution inference without patching at scales demonstrated up to 21 MP in this paper; a lightweight GNN classifies the reduced graph and an exact map lifts predictions to pixel resolution. One pipeline--identical architecture, features, loss, and GNN hyperparameters across all dataset--matches or exceeds domain-specific baselines on TTPLA (power lines), CrackSeg9k (pavement cracks), and SkyScapes Lane (aerial markings) on Dice, IoU, and Boundary F1 while reducing mask fragmentation by at least 4.6x relative to SLIC at matched inference.

cs.CV↗

arXiv2026

Transient Acceleration and Cross-Dissipation Interference in Fisher-Regularized Wasserstein Gradient Flows

We study transient nonequilibrium dynamics in Fisher-regularized Wasserstein gradient flows and identify a sign-changing cross-dissipation mechanism generated by the coupling between transport dissipation and Fisher-information geometry. Using the Ornstein--Uhlenbeck Fokker--Planck system as an analytically tractable setting, we derive an exact reduced variance dynamics on the Gaussian manifold, \[ \dot{u}=2(1-u)+\frac{\varepsilon}{u}, \] where $u(t)=σ^2(t)$ is the variance and $\varepsilon>0$ is the Fisher regularization strength. The reduced dynamics reveal distinct transient regimes induced by the interaction between transport relaxation and information-geometric curvature. The associated cross-dissipation term changes sign at the critical scale $σ=1$, separating cooperative acceleration for localized states with $σ<1$ from transient interference at larger variance scales. In the subcritical regime, Fisher curvature accelerates the descent of the baseline free energy; beyond the critical transition, it partially opposes the Ornstein--Uhlenbeck pullback and generates transient overshoot toward a displaced Fisher-regularized equilibrium. We also establish a bounded transient-acceleration-window result, showing that the cooperative acceleration phase has finite duration with an upper bound depending only on the Fisher regularization strength. Finite-difference simulations support the analytical predictions and suggest that qualitatively similar sign-transition behavior may persist beyond Gaussian closure for non-Gaussian initial conditions, including bimodal and Laplace distributions. Overall, the results provide a transient dynamical perspective on Fisher-regularized dissipative systems and show how information-geometric curvature can reorganize intermediate-time Wasserstein relaxation while preserving the globally dissipative structure of the flow.

cond-mat.stat-mech↗

arXiv2026

Empirical Characterization of Inference-Time Elicited Probability Transformations in Large Language Models

Large language models increasingly rely on inference-time procedures such as chain-of-thought reasoning, self-refinement, retrieval augmentation, and verifier-guided revision, yet the structure of elicited probability transformations under these procedures remains poorly understood. We study externally elicited probability assignments over candidate answers and observe recurring approximate log-ratio relationships: \[ \log \tilde q_t(i) = α_t \left( \log q_t(i) + \log b_t(i) \right) + c_t, \] where $q_t$ and $\tilde q_t$ are pre- and post-elicitation probabilities, $b_t$ is an externally constructed evidence signal, and $α_t$ is an empirical descriptor of the prompting configuration. Across 4,975 reasoning problems from GPQA Diamond, TheoremQA, MMLU-Pro, and ARC-Challenge, evaluated on multiple instruction-tuned model families, we observe approximate log-ratio relationships with mean $R^2 \approx 0.76$ over about $1.3 \times 10^5$ candidate-level observations. Coefficients vary across elicitation settings, but qualitatively similar relationships persist across evaluated conditions. Robustness analyses using alternative statistical representations, prompting configurations, held-out evaluation, and token-level log-probabilities suggest that the observed structure is not tied to one prompting procedure or probability estimation method. The main contribution is not the algebraic form itself, which is related to generalized Bayesian updating and probability-transformation frameworks, but the empirical observation that diverse inference-time prompting pipelines repeatedly exhibit reproducible log-ratio structure under controlled conditions. The framework provides a protocol-sensitive perspective for analyzing calibration, evidence amplification, uncertainty propagation, and interaction sensitivity in inference-time LLM pipelines.

cs.CL↗

arXiv2026

SEMIR: Semantic Minor-Induced Representation Learning on Graphs for Visual Segmentation

Segmenting small and sparse structures in large-scale images is fundamentally constrained by voxel-level, lattice-bound computation and extreme class imbalance -- dense, full-resolution inference scales poorly and forces most pipelines to rely on fixed regionization or downsampling, coupling computational cost to image resolution and attenuating boundary evidence precisely where minority structures are most informative. We introduce SEMIR (Semantic Minor-Induced Representation Learning), a representation framework that decouples inference from the native grid by learning a task-adapted, topology-preserving latent graph representation with exact decoding. SEMIR transforms the underlying grid graph into a compact, boundary-aligned graph minor through parameterized edge contraction, node deletion, and edge deletion, while preserving an exact lifting map from minor predictions to lattice labels. Minor construction is formalized as a few-shot structure learning problem that replaces hand-tuned preprocessing with a boundary-alignment objective: minor parameters are learned by maximizing agreement between predicted boundary elements and target-specific semantic edges under a boundary Dice criterion, and the induced minor is annotated with scale- and rotation-robust geometric and intensity descriptors and supports efficient region-level inference via message passing on a graph neural network (GNN) with relational edge features. We benchmark SEMIR on three tumor segmentation datasets -- BraTS 2021, KiTS23, and LiTS -- where targets exhibit high structural variability and distributional uncertainty. SEMIR yields consistent improvements in minority-structure Dice at practical runtime. More broadly, SEMIR establishes a framework for learning task-adapted, topology-preserving latent representations with exact decoding for high-resolution structured visual data.

cs.CV↗

arXiv2026

QuIC: A Training-Free Quantum Graph Embedding from Ideal Analysis to Practical Hardware Evaluation

We introduce QuIC, a training-free quantum graph embedding that maps graphs to sorted output distributions via a fixed parameterized circuit. In the ideal one-repetition setting, we prove that the resulting sorted distribution is permutation-invariant and injective on labeled graphs under an irrational-angle condition, yielding completeness on isomorphism classes for the ideal one-repetition exact-arithmetic embedding. We then use those ideal structural properties to motivate a practical embedding pipeline and study how much of that behavior survives under finite-shot estimation, truncation, realistic noise, transpilation, and hardware execution. The sorted distribution concentrates discriminative signal in a compact head, making fixed-length head truncation an effective practical operating point in the tested regimes. Under noise-model simulation, all tested graph pairs satisfied the study's operational separation criterion, including strongly regular graph pairs that are standard 2-WL stress tests and CFI families used as hard instances for fixed-k WL methods. A hardware study comprising 14,800 transpiled circuits across 37 CFI families on IBM Heron (ibm_fez, 156 qubits), including paired one- and two-repetition evaluations, reports empirical separation up to 66 qubits for the tested families under the reported execution protocol, identifies a device-dependent depth limit near 210-250 layers, and characterizes the current practical boundary of the method under the reported execution protocol.

quant-ph↗

arXiv2026

MATEX: Multi-scale Attention and Text-guided Explainability of Medical Vision-Language Models

We introduce MATEX (Multi-scale Attention and Text-guided Explainability), a novel framework that advances interpretability in medical vision-language models by incorporating anatomically informed spatial reasoning. MATEX synergistically combines multi-layer attention rollout, text-guided spatial priors, and layer consistency analysis to produce precise, stable, and clinically meaningful gradient attribution maps. By addressing key limitations of prior methods, such as spatial imprecision, lack of anatomical grounding, and limited attention granularity, MATEX enables more faithful and interpretable model explanations. Evaluated on the MS-CXR dataset, MATEX outperforms the state-of-the-art M2IB approach in both spatial precision and alignment with expert-annotated findings. These results highlight MATEX's potential to enhance trust and transparency in radiological AI applications.

cs.CV↗

arXiv2026

Predicting When to Trust Vision-Language Models for Spatial Reasoning

Vision-Language Models (VLMs) demonstrate impressive capabilities across multimodal tasks, yet exhibit systematic spatial reasoning failures, achieving only 49% (CLIP) to 54% (BLIP-2) accuracy on basic directional relationships. For safe deployment in robotics and autonomous systems, we need to predict when to trust VLM spatial predictions rather than accepting all outputs. We propose a vision-based confidence estimation framework that validates VLM predictions through independent geometric verification using object detection. Unlike text-based approaches relying on self-assessment, our method fuses four signals via gradient boosting: geometric alignment between VLM claims and coordinates, spatial ambiguity from overlap, detection quality, and VLM internal uncertainty. We achieve 0.674 AUROC on BLIP-2 (34.0% improvement over text-based baselines) and 0.583 AUROC on CLIP (16.1% improvement), generalizing across generative and classification architectures. Our framework enables selective prediction: at 60% target accuracy, we achieve 61.9% coverage versus 27.6% baseline (2.2x improvement) on BLIP-2. Feature analysis reveals vision-based signals contribute 87.4% of model importance versus 12.7% from VLM confidence, validating that external geometric verification outperforms self-assessment. We demonstrate reliable scene graph construction where confidence-based pruning improves precision from 52.1% to 78.3% while retaining 68.2% of edges.

cs.CV↗

arXiv2026

Clinical Knowledge Graph Construction and Evaluation with Multi-LLMs via Retrieval-Augmented Generation

Large language models (LLMs) offer new opportunities for constructing knowledge graphs (KGs) from unstructured clinical narratives. However, existing approaches often rely on structured inputs and lack robust validation of factual accuracy and semantic consistency, limitations that are especially problematic in oncology. We introduce an end-to-end framework for clinical KG construction and evaluation directly from free text using multi-agent prompting and a schema-constrained Retrieval-Augmented Generation (KG-RAG) strategy. Our pipeline integrates (1) prompt-driven entity, attribute, and relation extraction; (2) entropy-based uncertainty scoring; (3) ontology-aligned RDF/OWL schema generation; and (4) multi-LLM consensus validation for hallucination detection and semantic refinement. Beyond static graph construction, the framework supports continuous refinement and self-supervised evaluation, enabling iterative improvement of graph quality. Applied to two oncology cohorts (PDAC and BRCA), our method produces interpretable, SPARQL-compatible, and clinically grounded knowledge graphs without relying on gold-standard annotations. Experimental results demonstrate consistent gains in precision, relevance, and ontology compliance over baseline methods.

cs.AI↗

arXiv2025

DGTEN: A Robust Deep Gaussian based Graph Neural Network for Dynamic Trust Evaluation with Uncertainty-Quantification Support

Dynamic trust evaluation in large, rapidly evolving graphs demands models that capture changing relationships, express calibrated confidence, and resist adversarial manipulation. DGTEN (Deep Gaussian-Based Trust Evaluation Network) introduces a unified graph-based framework that does all three by combining uncertainty-aware message passing, expressive temporal modeling, and built-in defenses against trust-targeted attacks. It represents nodes and edges as Gaussian distributions so that both semantic signals and epistemic uncertainty propagate through the graph neural network, enabling risk-aware trust decisions rather than overconfident guesses. To track how trust evolves, it layers hybrid absolute-Gaussian-hourglass positional encoding with Kolmogorov-Arnold network-based unbiased multi-head attention, then applies an ordinary differential equation-based residual learning module to jointly model abrupt shifts and smooth trends. Robust adaptive ensemble coefficient analysis prunes or down-weights suspicious interactions using complementary cosine and Jaccard similarity, curbing reputation laundering, sabotage, and on-off attacks. On two signed Bitcoin trust networks, DGTEN delivers standout gains where it matters most: in single-timeslot prediction on Bitcoin-OTC, it improves MCC by +12.34% over the best dynamic baseline; in the cold-start scenario on Bitcoin-Alpha, it achieves a +25.00% MCC improvement, the largest across all tasks and datasets; while under adversarial on-off attacks, it surpasses the baseline by up to +10.23% MCC. These results endorse the unified DGTEN framework.

cs.LG↗

arXiv2025

PathFormer: A Transformer with 3D Grid Constraints for Digital Twin Robot-Arm Trajectory Generation

Robotic arms require precise, task-aware trajectory planning, yet sequence models that ignore motion structure often yield invalid or inefficient executions. We present a Path-based Transformer that encodes robot motion with a 3-grid (where/what/when) representation and constraint-masked decoding, enforcing lattice-adjacent moves and workspace bounds while reasoning over task graphs and action order. Trained on 53,755 trajectories (80% train / 20% validation), the model aligns closely with ground truth -- 89.44% stepwise accuracy, 93.32% precision, 89.44% recall, and 90.40% F1 -- with 99.99% of paths legal by construction. Compiled to motor primitives on an xArm Lite 6 with a depth-camera digital twin, it attains up to 97.5% reach and 92.5% pick success in controlled tests, and 86.7% end-to-end success across 60 language-specified tasks in cluttered scenes, absorbing slips and occlusions via local re-grounding without global re-planning. These results show that path-structured representations enable Transformers to generate accurate, reliable, and interpretable robot trajectories, bridging graph-based planning and sequence-based learning and providing a practical foundation for general-purpose manipulation and sim-to-real transfer.

cs.RO↗

arXiv2025

GRIP: A Unified Framework for Grid-Based Relay and Co-Occurrence-Aware Planning in Dynamic Environments

Robots navigating dynamic, cluttered, and semantically complex environments must integrate perception, symbolic reasoning, and spatial planning to generalize across diverse layouts and object categories. Existing methods often rely on static priors or limited memory, constraining adaptability under partial observability and semantic ambiguity. We present GRIP, Grid-based Relay with Intermediate Planning, a unified, modular framework with three scalable variants: GRIP-L (Lightweight), optimized for symbolic navigation via semantic occupancy grids; GRIP-F (Full), supporting multi-hop anchor chaining and LLM-based introspection; and GRIP-R (Real-World), enabling physical robot deployment under perceptual uncertainty. GRIP integrates dynamic 2D grid construction, open-vocabulary object grounding, co-occurrence-aware symbolic planning, and hybrid policy execution using behavioral cloning, D* search, and grid-conditioned control. Empirical results on AI2-THOR and RoboTHOR benchmarks show that GRIP achieves up to 9.6% higher success rates and over $2\times$ improvement in path efficiency (SPL and SAE) on long-horizon tasks. Qualitative analyses reveal interpretable symbolic plans in ambiguous scenes. Real-world deployment on a Jetbot further validates GRIP's generalization under sensor noise and environmental variation. These results position GRIP as a robust, scalable, and explainable framework bridging simulation and real-world navigation.

cs.RO↗

arXiv2025

Multi-Modal Interpretability for Enhanced Localization in Vision-Language Models

Recent advances in vision-language models have significantly expanded the frontiers of automated image analysis. However, applying these models in safety-critical contexts remains challenging due to the complex relationships between objects, subtle visual cues, and the heightened demand for transparency and reliability. This paper presents the Multi-Modal Explainable Learning (MMEL) framework, designed to enhance the interpretability of vision-language models while maintaining high performance. Building upon prior work in gradient-based explanations for transformer architectures (Grad-eclip), MMEL introduces a novel Hierarchical Semantic Relationship Module that enhances model interpretability through multi-scale feature processing, adaptive attention weighting, and cross-modal alignment. Our approach processes features at multiple semantic levels to capture relationships between image regions at different granularities, applying learnable layer-specific weights to balance contributions across the model's depth. This results in more comprehensive visual explanations that highlight both primary objects and their contextual relationships with improved precision. Through extensive experiments on standard datasets, we demonstrate that by incorporating semantic relationship information into gradient-based attribution maps, MMEL produces more focused and contextually aware visualizations that better reflect how vision-language models process complex scenes. The MMEL framework generalizes across various domains, offering valuable insights into model decisions for applications requiring high interpretability and reliability.

cs.CV↗

arXiv2025

Re-Imagining Multimodal Instruction Tuning: A Representation View

Multimodal instruction tuning has proven to be an effective strategy for achieving zero-shot generalization by fine-tuning pre-trained Large Multimodal Models (LMMs) with instruction-following data. However, as the scale of LMMs continues to grow, fully fine-tuning these models has become highly parameter-intensive. Although Parameter-Efficient Fine-Tuning (PEFT) methods have been introduced to reduce the number of tunable parameters, a significant performance gap remains compared to full fine-tuning. Furthermore, existing PEFT approaches are often highly parameterized, making them difficult to interpret and control. In light of this, we introduce Multimodal Representation Tuning (MRT), a novel approach that focuses on directly editing semantically rich multimodal representations to achieve strong performance and provide intuitive control over LMMs. Empirical results show that our method surpasses current state-of-the-art baselines with significant performance gains (e.g., 1580.40 MME score) while requiring substantially fewer tunable parameters (e.g., 0.03% parameters). Additionally, we conduct experiments on editing instrumental tokens within multimodal representations, demonstrating that direct manipulation of these representations enables simple yet effective control over network behavior.

cs.LG↗

arXiv2023

OCU-Net: A Novel U-Net Architecture for Enhanced Oral Cancer Segmentation

Accurate detection of oral cancer is crucial for improving patient outcomes. However, the field faces two key challenges: the scarcity of deep learning-based image segmentation research specifically targeting oral cancer and the lack of annotated data. Our study proposes OCU-Net, a pioneering U-Net image segmentation architecture exclusively designed to detect oral cancer in hematoxylin and eosin (H&E) stained image datasets. OCU-Net incorporates advanced deep learning modules, such as the Channel and Spatial Attention Fusion (CSAF) module, a novel and innovative feature that emphasizes important channel and spatial areas in H&E images while exploring contextual information. In addition, OCU-Net integrates other innovative components such as Squeeze-and-Excite (SE) attention module, Atrous Spatial Pyramid Pooling (ASPP) module, residual blocks, and multi-scale fusion. The incorporation of these modules showed superior performance for oral cancer segmentation for two datasets used in this research. Furthermore, we effectively utilized the efficient ImageNet pre-trained MobileNet-V2 model as a backbone of our OCU-Net to create OCU-Netm, an enhanced version achieving state-of-the-art results. Comprehensive evaluation demonstrates that OCU-Net and OCU-Netm outperformed existing segmentation methods, highlighting their precision in identifying cancer cells in H&E images from OCDC and ORCA datasets.

eess.IV↗

arXiv2021

EARLIN: Early Out-of-Distribution Detection for Resource-efficient Collaborative Inference

Collaborative inference enables resource-constrained edge devices to make inferences by uploading inputs (e.g., images) to a server (i.e., cloud) where the heavy deep learning models run. While this setup works cost-effectively for successful inferences, it severely underperforms when the model faces input samples on which the model was not trained (known as Out-of-Distribution (OOD) samples). If the edge devices could, at least, detect that an input sample is an OOD, that could potentially save communication and computation resources by not uploading those inputs to the server for inference workload. In this paper, we propose a novel lightweight OOD detection approach that mines important features from the shallow layers of a pretrained CNN model and detects an input sample as ID (In-Distribution) or OOD based on a distance function defined on the reduced feature space. Our technique (a) works on pretrained models without any retraining of those models, and (b) does not expose itself to any OOD dataset (all detection parameters are obtained from the ID training dataset). To this end, we develop EARLIN (EARLy OOD detection for Collaborative INference) that takes a pretrained model and partitions the model at the OOD detection layer and deploys the considerably small OOD part on an edge device and the rest on the cloud. By experimenting using real datasets and a prototype implementation, we show that our technique achieves better results than other approaches in terms of overall accuracy and cost when tested against popular OOD datasets on top of popular deep learning models pretrained on benchmark datasets.

cs.CV↗

arXiv2020

SCAT: Second Chance Autoencoder for Textual Data

We present a k-competitive learning approach for textual autoencoders named Second Chance Autoencoder (SCAT). SCAT selects the $k$ largest and smallest positive activations as the winner neurons, which gain the activation values of the loser neurons during the learning process, and thus focus on retrieving well-representative features for topics. Our experiments show that SCAT achieves outstanding performance in classification, topic modeling, and document visualization compared to LDA, K-Sparse, NVCTM, and KATE.

cs.CL↗

arXiv2020

Link Prediction for Temporally Consistent Networks

Dynamic networks have intrinsic structural, computational, and multidisciplinary advantages. Link prediction estimates the next relationship in dynamic networks. However, in the current link prediction approaches, only bipartite or non-bipartite but homogeneous networks are considered. The use of adjacency matrix to represent dynamically evolving networks limits the ability to analytically learn from heterogeneous, sparse, or forming networks. In the case of a heterogeneous network, modeling all network states using a binary-valued matrix can be difficult. On the other hand, sparse or currently forming networks have many missing edges, which are represented as zeros, thus introducing class imbalance or noise. We propose a time-parameterized matrix (TP-matrix) and empirically demonstrate its effectiveness in non-bipartite, heterogeneous networks. In addition, we propose a predictive influence index as a measure of a node's boosting or diminishing predictive influence using backward and forward-looking maximization over the temporal space of the n-degree neighborhood. We further propose a new method of canonically representing heterogeneous time-evolving activities as a temporally parameterized network model (TPNM). The new method robustly enables activities to be represented as a form of a network, thus potentially inspiring new link prediction applications, including intelligent business process management systems and context-aware workflow engines. We evaluated our model on four datasets of different network systems. We present results that show the proposed model is more effective in capturing and retaining temporal relationships in dynamically evolving networks. We also show that our model performed better than state-of-the-art link prediction benchmark results for networks that are sensitive to temporal evolution.

cs.SI↗

arXiv2020

CRL: Class Representative Learning for Image Classification

Building robust and real-time classifiers with diverse datasets are one of the most significant challenges to deep learning researchers. It is because there is a considerable gap between a model built with training (seen) data and real (unseen) data in applications. Recent works including Zero-Shot Learning (ZSL), have attempted to deal with this problem of overcoming the apparent gap through transfer learning. In this paper, we propose a novel model, called Class Representative Learning Model (CRL), that can be especially effective in image classification influenced by ZSL. In the CRL model, first, the learning step is to build class representatives to represent classes in datasets by aggregating prominent features extracted from a Convolutional Neural Network (CNN). Second, the inferencing step in CRL is to match between the class representatives and new data. The proposed CRL model demonstrated superior performance compared to the current state-of-the-art research in ZSL and mobile deep learning. The proposed CRL model has been implemented and evaluated in a parallel environment, using Apache Spark, for both distributed learning and recognition. An extensive experimental study on the benchmark datasets, ImageNet-1K, CalTech-101, CalTech-256, CIFAR-100, shows that CRL can build a class distribution model with drastic improvement in learning and recognition performance without sacrificing accuracy compared to the state-of-the-art performances in image classification.

cs.CV↗