Zhiyuan Liang

DD-Ranking: Rethinking the Evaluation of Dataset Distillation

Towards Personalized Parameter Generation via Data-Conditioned Mapping

Faster Vision Mamba is Rebuilt in Minutes via Merged Token Re-training

Drag-and-Drop LLMs: Zero-Shot Prompt-to-Weights

Pruning-Robust Mamba with Asymmetric Multi-Scale Scanning Paths

CREAM: Consistency Regularized Self-Rewarding Language Models

REPA Works Until It Doesn’t: Early-Stopped, Holistic Alignment Supercharges Diffusion Training