4 papers
GVCC: Zero-Shot Video Compression via Codebook-Driven Stochastic Rectified Flow
Ziyue Zeng, Xun Su, Haoyuan Liu +3
At ultra-low bitrates, high-fidelity reconstruction requires sampling plausible videos from the posterior rather than regressing to oversmoothed conditional means. We propose Gener…
FEPLB: Exploiting Copy Engines for Nearly Free MoE Load Balancing in Distributed Training
Shuyao Qi, Haoyuan Liu, Shizhen Zhao
Fine-grained, per-micro-batch load balancing is essential for efficient Mixture-of-Experts (MoE) training, yet every prior dynamic scheduling scheme pays for it with extra communic…
InterpIoU: Rethinking Bounding Box Regression with Interpolation-Based IoU Optimization
Haoyuan Liu, Hiroshi Watanabe
Bounding box regression (BBR) is fundamental to object detection, where the regression loss is crucial for accurate localization. Existing IoU-based losses often incorporate handcr…
Time Step Generating: A Universal Synthesized Deepfake Image Detector
Ziyue Zeng, Haoyuan Liu, Dingjie Peng +2
Currently, high-fidelity text-to-image models are developed in an accelerating pace. Among them, Diffusion Models have led to a remarkable improvement in the quality of image gener…