Irminsul: MLA-Native Position-Independent Caching for Agentic LLM Serving

Documenti analoghi

• Leyline: KV Cache Directives for Agentic Inference
  di: Ma, Bole, et al.
  Pubblicazione: (2026)
• Move the Query, Not the Cache: Characterizing Cross-Instance Latent Attention Redistribution Across GPU Fabrics
  di: Ma, Bole, et al.
  Pubblicazione: (2026)
• Diagnosing Overhead in Dispatch Operations: Cross-architecture Observatory
  di: Ma, Bole, et al.
  Pubblicazione: (2026)
• The Illusion of Power Capping in LLM Decode: A Phase-Aware Energy Characterisation Across Attention Architectures
  di: Ma, Bole, et al.
  Pubblicazione: (2026)
• FlashMLA-ETAP: Efficient Transpose Attention Pipeline for Accelerating MLA Inference on NVIDIA H20 GPUs
  di: Dege, Pengcuo, et al.
  Pubblicazione: (2025)