Z Lab ORG

DFlash: Block Diffusion for Flash Speculative Decoding

[ICLR 2026] ParoQuant: Pairwise Rotation Quantization for Efficient Reasoning LLM Inference

[ICML 2025] SparseLoRA: Accelerating LLM Fine-Tuning with Contextual Sparsity

Fast, memory-efficient attention column reduction (e.g., sum, mean, max)