Zhen Qin

Householder-Diagonalized Linear Attention (HDLA): Utilizing Enhanced Decay Mechanism for Efficient Sequence Modeling

Accelerate Autoregressive Normalizing Flows Sampling with GS-Jacobi Iteration

An Efficient Framework for Length Extension via Dynamically Growing Positional Embedding and Correlation-Aware Routing Attention

Discovering Architectures via an Evolutionary Agentic Framework

Group Representational Position Encoding

Tensor Product Attention Is All You Need

Elucidating the Design Space of Decay in Linear Attention

Tensor Product Attention Is All You Need

MVA: Linear Attention with High-order Query-Keys Integration and Multi-level Vocabulary Decomposition

You Only Scan Once: Efficient Multi-dimension Sequential Modeling with LightNet

Linear Attention Sequence Parallelism

FlashSampling: Fast and Memory-Efficient Exact Sampling with Group-Gumbel-Max