SplitZip: Ultra Fast Lossless KV Compression for Disaggregated LLM Serving
cs.DC, cs.AI, cs.LG
Submitted: 2026-05-03
Updated: 2026-11-13
Code: https://github.com/Intelligent-Microsystems-Lab/SplitZip
Terminology
Sources
- Quad Length Codes for Lossless Compression of e4m3
- Single-Stage Huffman Encoder for ML Compression
- Longformer: The Long-Document Transformer
- Evaluating Large Language Models Trained on Code
- Training Verifiers to Solve Math Word Problems
- ZipNN: Lossless Compression for AI Models
- FlowKV: A Disaggregated Inference Framework with Low-Latency KV Cache Transfer and Load-Aware Scheduling
- A High-Throughput GPU Framework for Adaptive Lossless Compression of Floating-Point Data
- Pointer Sentinel Mixture Models
- The Llama 3 Herd of Models
- Splitwise: Efficient generative LLM inference using phase splitting
- Prefill-as-a-Service: KVCache of Next-Generation Models Could Go Cross-Datacenter
- KVLinC : KV Cache Quantization with Hadamard Rotation and Linear Correction
- ZipLLM: Efficient LLM Storage via Model-Aware Synergistic Data Deduplication and Compression
- Qwen3 Technical Report
- Huff-LLM: End-to-End Lossless Compression for Efficient LLM Inference
- Head-Driven Phrase Structure Grammar Parsing on Penn Treebank
Related papers
- BatchGen: An Architecture for Scalable and Efficient Batch Inference
- Seamless Reconfiguration for DAG-BFT (Technical Report)
- ELDR: Expert-Locality-Aware Decode Routing for PD-Disaggregated MoE Serving
- iScheduler: Reinforcement Learning-Driven Continual Optimization for Large-Scale Resource Investment Problems
- TIRx: A Unified Tile Primitive ML Compiler Across Heterogeneous Backends
- AgentWeave: Efficient Distributed Agent Serving via Flow Decomposition