AdaSplash-2 Speeds Up Sparse Attention for LLMs at ICML 2026

#ICML2026 AdaSplash2 accelerates sparse attention, α-entmax and introduced a histogram-based initialization for attention weights. AdaSplash2 achieves substantially faster computation than FlashAttention-2. https://x.com/lileics/status/2075051879130632695/photo/1 https://twitter.com/lileics/status/2075051223074378051