HATA: Trainable and Hardware-Efficient Hash-Aware Top-k Attention for Scalable Large Model Inference

Documenti analoghi

• Efficient Long-Context LLM Inference via KV Cache Clustering
  di: Hu, Jie, et al.
  Pubblicazione: (2025)
• AdaCluster: Adaptive Query-Key Clustering for Sparse Attention in Video Generation
  di: Tan, Haoyue, et al.
  Pubblicazione: (2026)
• A Mathematical Theory of Top-$k$ Sparse Attention via Total Variation Distance
  di: Tzachristas, Georgios, et al.
  Pubblicazione: (2025)
• XL3M: A Training-free Framework for LLM Length Extension Based on Segment-wise Inference
  di: Wang, Shengnan, et al.
  Pubblicazione: (2024)
• LiteCache: A Query Similarity-Driven, GPU-Centric KVCache Subsystem for Efficient LLM Inference
  di: Yi, Jiawei, et al.
  Pubblicazione: (2025)