Sparse Attention with Persistent State Machines – High‑Sparsity LLM Accelerator(zenodo.org)3 points by yusuke_esaka 3 days ago | 0 comments