What Remains Normal? Clean Images Miss Useful Near-Defect Normal Patches for Anomaly Detection
Under review at ICLR, 2026.
BoundarySupport exposes missing near-defect normal evidence under fixed memory and reconstruction settings.
Publications
My papers and preprints span anomaly detection, medical AI, multimodal learning, and generative AI. Google Scholar
2026
What Memory Composition Does Not Tell Us About Anomaly Detection
Under review at ICLR, 2026.
CLEANCON separates memory purity from memory utility through controlled candidate-image eligibility.
ProCon: Projection-Consistency Memory for Training-Free Anomaly Detection
Under review, 2026.
Projection-consistency memory selection for frozen anomaly detectors.
GCR: Geometry-Consistent Routing for Task-Agnostic Continual Anomaly Detection
Under review, 2026.
Cross-head routing for continual anomaly detection without retraining.
StructCore: Structure-Aware Image-Level Scoring for Training-Free Unsupervised Anomaly Detection
Under review at WACV, 2026.
Structure-aware calibration from anomaly maps to image scores.
MMIR-TCM: Memory-Integrated Multimodal Inference and Retrieval for TCM Clinical Decision Support
Major revision at IEEE JBHI, 2026.
Memory-integrated retrieval for multimodal clinical decision support.
ViGen: Video-based Generation with GRPO for Dynamic Image Editing
Under review at AAAI, 2026.
Video-based generation with GRPO for dynamic image editing.
2025
Memory-SAM: Human-Prompt-Free Tongue Segmentation via Retrieval-to-Prompt
Under review at AAAI 2027, 2025.
Foreground-minus-background DINOv3 contrast generates automatic SAM2 prompts; SM-Tongue releases 2,155 image-mask pairs. Earlier versions averaged 24/28 at ICASSP 2026 and 4.0/5.0 at MICCAI 2026.
Auditable Context-Aware HFMD Forecasting with Structured LLM Agents
Under review at IEEE BIBM, 2025.
Structured LLM agents for auditable, context-aware HFMD forecasting.
ViTCM-LLM: A Multimodal RAG Framework for Advanced TCM Clinical Decision Support
2025 IEEE International Conference on Bioinformatics and Biomedicine (BIBM), pp. 3894-3899, 2025.
Multimodal RAG and evaluation for TCM clinical decision support.
Pre-operative T-stage discrimination in gallbladder cancer using machine learning and DeepSeek-R1
Frontiers in Oncology, 15, Article 1613462, 2025.
Clinical evaluation across 232 patients with bootstrap confidence intervals.
2024
Grid-augmented vision: A simple yet effective approach for enhanced spatial understanding in multi-modal agents
arXiv:2411.18270 [cs.CV], 2024.
Grid augmentation for spatial understanding in multimodal agents.