1 paper
Piotr Nawrot, Robert Li, Renjie Huang +3
Sparse attention offers a promising strategy to extend long-context capabilities in Transformer LLMs, yet its efficiency-accuracy trade-offs remain unclear due to the lack of compr…