Yiwu Yao

FreeKV: Boosting KV Cache Retrieval for Efficient LLM Inference

Astraea: A Token-wise Acceleration Framework for Video Diffusion Transformers

Compact Attention: Exploiting Structured Spatio-Temporal Sparsity for Fast Video Generation

DartQuant: Efficient Rotational Distribution Calibration for LLM Quantization

RazorAttention: Efficient KV Cache Compression Through Retrieval Heads

Dynamic Low-Rank Sparse Adaptation for Large Language Models

KVTuner: Sensitivity-Aware Layer-Wise Mixed-Precision KV Cache Quantization for Efficient and Nearly Lossless LLM Inference

Extreme composite compression of large language models through joint optimization