🏢 University of Hong Kong

MimicBrush: a novel image editing approach using reference imitation for intuitive zero-shot edits.

CLIP’s robustness to long-tailed pre-training data stems from its dynamic classification task and descriptive language supervision, offering transferable insights for improving model generalizability.

Task-oriented time series imputation is revolutionized! This research introduces a novel approach that efficiently assesses imputation strategies based on downstream task performance without costly mo…

SyncVIS: A new framework for video instance segmentation achieves state-of-the-art results by synchronously modeling video and frame-level information, overcoming limitations of asynchronous approache…

Stacking Your Transformers accelerates LLM pre-training by leveraging smaller, pre-trained models to efficiently train larger ones, achieving significant speedups and improved performance.

Researchers introduce Video Gaussian Representation (VGR) for versatile video processing, embedding videos into explicit 3D Gaussians for intuitive motion and appearance modeling.

ArithTreeRL, a novel reinforcement learning approach, generates optimized arithmetic tree structures for adders and multipliers, significantly improving computational efficiency and reducing hardware …

MAGIS: A novel LLM-based multi-agent framework significantly boosts GitHub issue resolution by leveraging agent collaboration for planning and coding, achieving an eight-fold performance increase comp…

KptLLM: A novel multimodal model leverages LLMs for superior keypoint comprehension, outperforming existing methods in various benchmarks.

Multi-head transformers utilize distinct attention patterns across layers—multiple heads are essential for initial data preprocessing, while a single head suffices for subsequent optimization steps, o…

HHD-GP leverages Helmholtz-Hodge decomposition within Gaussian Processes to learn physically meaningful components of dynamical systems, enhancing prediction accuracy and interpretability.

EFFI-LEARNER: A novel self-optimization framework dramatically improves the efficiency of LLM-generated code by iteratively refining code based on execution profiles.

AutoPSV automates process annotation for LLMs, improving reasoning by detecting confidence shifts in reasoning steps, thus efficiently enhancing model performance.