SCAD: Super-Class-Aware Debiasing for Long-Tailed Semi-Supervised Learning

SCAD: Super-Class-Aware Debiasing for Long-Tailed Semi-Supervised LearningSunguk Jang*, Jinwoo Jeon*, Byung-Jun Lee (*: equal contribution)In long-tailed semi-supervised learning (LTSSL), pseudolabeling often creates a vicious cycle of bias amplification, a problem that recent state-of-the-art metho...

Soft Equivariance Regularization for Invariant Self-Supervised Learning

Soft Equivariance Regularization for Invariant Self-Supervised LearningJoohyung Lee, Changhun Kim, Hyunsu Kim, Kwanhyung Lee, Juho LeeA central principle in self-supervised learning (SSL) is to learn data representations that are invariant to semantic-preserving transformations eg, image representat...

Delta-XAI: A Unified Framework for Explaining Prediction Changes in Online Time Series Monitoring

Delta-XAI: A Unified Framework for Explaining Prediction Changes in Online Time Series MonitoringChanghun Kim, Yechan Mun, Hyeongwon Jang, Eunseo Lee, Sangchul Hahn, Eunho YangExplaining online time series monitoring models is crucial across sensitive domains such as healthcare and finance, where te...

Test-Time Ensemble via Linear Mode Connectivity: A Path to Better Adaptation

ICLR 2025 Test-Time Ensemble via Linear Mode Connectivity: A Path to Better Adaptation  Byungjai Kim, Chanho Ahn, Wissam J. Baddar, Kikyung Kim, HUIJIN LEE, Saehyun Ahn, Seungju Han, Sungjoo Suh, Eunho Yang  Test-time adaptation updates pretrained models on the fly to handle...

LANTERN++: Enhanced Relaxed Speculative Decoding with Static Tree Drafting for Visual Auto-regressive Models

ICLR 2025 Workshop on Scalable Optimization for Efficient and Adaptive Foundation Models (ICLRW-SCOPE) (Oral) LANTERN++: Enhanced Relaxed Speculative Decoding with Static Tree Drafting for Visual Auto-regressive Models Sihwan Park, Doohyuk Jang, Sung-Yub Kim, Souvik Kundu, Eunho Yang&...

TIMING: Temporality-Aware Integrated Gradients for Time Series Explanation

ICLR 2025 Workshop on XAI4Science: From Understanding Model Behavior to Discovering New Scientific Knowledge (ICLRW-XAI4Science) TIMING: Temporality-Aware Integrated Gradients for Time Series Explanation Hyeongwon Jang, Changhun Kim, Eunho Yang  Recent explainable artif...

Token-Supervised Value Models for Enhancing Mathematical Problem-Solving Capabilities of Large Language Models

ICLR 2025 Token-Supervised Value Modelsfor Enhancing Mathematical Problem-Solving Capabilities of Large Language Models  Jung Hyun Lee, June Yong Yang, Byeongho Heo, Dongyoon Han, Kyungsu Kim, Eunho Yang, Kang Min Yoo  With the rapid advancement of test-time compute sea...

ReBind: Enhancing Ground-state Molecular Conformation Prediction via Force-Based Graph Rewiring

ICLR 2025 ReBind: Enhancing Ground-state Molecular Conformation Prediction via Force-Based Graph Rewiring Taewon Kim, Hyunjin Seo, Sungsoo Ahn, Eunho Yang  Predicting the ground-state 3D molecular conformations from 2D molecular graphs is critical in computational chemistry due t...

LANTERN: Accelerating Visual Autoregressive Models with Relaxed Speculative Decoding

ICLR 2025 LANTERN: Accelerating Visual Autoregressive Models with Relaxed Speculative Decoding Doohyuk Jang, Sihwan Park, June Yong Yang, Yeonsung Jung, Jihun Yun, Souvik Kundu, Sung-Yub Kim, Eunho Yang  Auto-Regressive (AR) models have recently gained prominence in image generat...

Divide and Translate: Compositional First-Order Logic Translation and Verification for Complex Logical Reasoning

ICLR 2025 Divide and Translate: Compositional First-Order Logic Translation and Verification for Complex Logical Reasoning Hyun Ryu, Gyeongman Kim, Hyemin S. Lee, Eunho Yang  Complex logical reasoning tasks require a long sequence of reasoning, which a large language model (...