RedKnot: Efficient Long-Context LLM Serving with Head-Aware KV Reuse and SegPagedAttention
cs.AI
Submitted: 2026-06-04
Updated: 2026-09-21
Code: https://github.com/deepseek-ai/Flas
License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Terminology
Sources
- LongBench: A Bilingual, Multitask Benchmark for Long Context Understanding
- The Llama 3 Herd of Models
- Mixture of Attention Spans: Optimizing LLM Inference Efficiency with Heterogeneous Sliding-Window Lengths
- Retrieval-Augmented Generation for Large Language Models: A Survey
- Mistral 7B
- CompressKV: Semantic Retrieval Heads Know What Tokens are Not Important Before Generation
- Llama 2: Open Foundation and Fine-Tuned Chat Models
- FlashAttention-3: Fast and Accurate Attention with Asynchrony and Low-precision
- ProphetKV: User-Query-Driven Selective Recomputation for Efficient KV Cache Reuse in Retrieval-Augmented Generation
- Fast Transformer Decoding: One Write-Head is All You Need
- LongGenBench: Benchmarking Long-Form Generation in Long Context LLMs
- Qwen3 Technical Report
Related papers
- MAVEN-T: Reinforced Heterogeneous Distillation for Real-Time Multi-Agent Trajectory Prediction
- Model Discovery Agent: LLM-assisted Bayesian experiment design for data-efficient discovery of mechanistic world models
- The Clinician's Veto: Navigating Trust, Liability, and Uncertainty in Autonomous AI Prescribing
- MindHelper: Closed-Loop Embodied Mental-State Reasoning for Precision Intervention
- Incumbent Advantage: Brand Bias and Cognitive Manipulation Dynamics in LLM Recommendation Systems
- VSAL: A Vision Solver with Adaptive Layouts for Graph Property Detection