IndexCache: Fixing Sparse Attention's Hidden O(L²) Problem

IndexCache identifies that DeepSeek Sparse Attention reduces core attention to O(Lk) but its lightning indexer retains O(L²) complexity. By exploiting high cross-layer index stability to reuse token selection indices, IndexCache delivers significant wall-clock throughput improvements for long-context inference without accuracy loss. Drop-in optimization for any production system using DeepSeek-style sparse attention.