Institution profile

Pohang University of Science and Technology

Academic institutionasia · kr
Official website
Research library524linked papers
Opportunities0open roles
Selected work

Representative Papers

Nonparametric estimation of a factorizable density using diffusion models

Jan 03, 2025

To address the curse of dimensionality in high-dimensional nonparametric density estimation, this paper considers densities exhibiting a low-dimensional factorized structure—i.e., statistical independence across variable groups. We propose diffusion models as implicit density estimators and, for the first time within a statistical framework, establish that under the factorization assumption, the resulting estimator achieves a dimension-free minimax-optimal convergence rate in total variation distance—thereby circumventing the curse of dimensionality. To explicitly encode structural priors, we design a sparse weight-sharing neural network architecture that adaptively models low-dimensional components. Theoretical analysis confirms the improved statistical efficiency, while empirical results demonstrate superior estimation accuracy and enhanced interpretability in high-dimensional sparse settings—all without sacrificing the flexibility inherent to nonparametric methods.

4 citations1 influentialRead paper

ANUBIS: Skeleton Action Recognition Dataset, Review, and Benchmark

May 04, 2022arXiv.org

Existing 3D skeleton-based action recognition research suffers from fragmented representation taxonomies and evaluation protocols misaligned with real-world scenarios; moreover, mainstream datasets lack critical dimensions—including rear-view perspectives, multi-person interactions, fine-grained or violent actions, and pandemic-era behaviors. To address these gaps, we propose a four-dimensional taxonomy (dataset design, spatial modeling, temporal modeling, and signal enhancement) and introduce ANUBIS: the first large-scale, multi-view 3D skeleton dataset explicitly designed for realistic challenges. ANUBIS features rear-view captures, 101 action classes (including 21 pandemic-related behaviors), and standardized recordings from 128 participants using Azure Kinect’s multi-sensor fusion. We further establish a unified benchmark framework, enabling reproducible evaluation of 12 state-of-the-art models. Our analysis identifies temporal modeling capacity and signal robustness as the primary bottlenecks limiting current performance.

4 citationsRead paper

Efficient Real-World Deblurring using Single Images: AIM 2025 Challenge Report

Oct 14, 2025

This work addresses efficient single-image deblurring for real-world scenarios under strict lightweight constraints: <5 M parameters and <200 GMACs. Leveraging our newly introduced RSBlur dataset—collected via a dual-camera setup and containing paired sharp-blurry images—we propose a lightweight convolutional backbone, a channel-spatial collaborative attention module, and a multi-stage feature recalibration mechanism to achieve high-fidelity restoration with minimal computational overhead. On the RSBlur test set, our method achieves 31.1298 dB PSNR, setting the new state-of-the-art among all approaches satisfying the specified efficiency constraints. Its feasibility and scalability are further validated by four independent participating teams in a benchmarking challenge. To the best of our knowledge, this is the first study to systematically define, construct, and empirically validate a practical lightweight benchmark for real-world image deblurring—bridging the gap between algorithmic performance and edge-device deployment.

3 citationsRead paper

Optimizing LLM Inference for Database Systems: Cost-Aware Scheduling for Concurrent Requests

Nov 12, 2024

To address performance bottlenecks—such as high GPU resource overhead, low throughput, and elevated latency—caused by concurrent requests in database-embedded LLM inference, this work pioneers the adaptation of database multi-query optimization principles to LLM inference systems. We propose a fine-grained GPU memory cost model that jointly characterizes VRAM occupancy and memory bandwidth constraints, along with a cooperative scheduling strategy integrating batch-aware scheduling and dynamic priority reordering. Evaluated against baseline approaches, our method achieves up to 47% higher inference throughput, 32% lower average latency, and 32% reduced GPU resource consumption—all without compromising model accuracy. The core contribution lies in cross-paradigm transfer of database optimization techniques to LLM inference, enabling resource modeling, schedulability optimization, and performance predictability—thereby establishing a principled foundation for efficient, scalable, and predictable LLM inference within database systems.

3 citationsRead paper

LILaC: Late Interacting in Layered Component Graph for Open-domain Multimodal Multihop Retrieval

Feb 04, 2026Conference on Empirical Methods in Natural Language Processing

This work addresses the limitations of fixed single-granularity retrieval units in open-domain multimodal multi-hop retrieval, which introduce noise and struggle to capture cross-document semantic relationships. To overcome these challenges, the authors propose a hierarchical component graph structure that jointly models multimodal information at both coarse and fine granularities. They further design an edge-based late-interaction subgraph retrieval mechanism that first performs coarse-grained candidate filtering followed by fine-grained reasoning. This approach achieves state-of-the-art retrieval performance across all five benchmark datasets without requiring additional fine-tuning, effectively balancing computational efficiency with multi-hop reasoning accuracy.

2 citations1 influentialRead paper
Recent publications

Latest Papers