repoGitHubTrust 82 · PrimaryPublished 5d agoLive · 4m ago
jagmarques/nexusquant
Training-free KV cache compression via E8 lattice VQ. 2-bit KV that preserves retrieval (30/30 NIAH vs TurboQuant 0/30). Calibration-free, 9 architectures validated.
Lineage graph
Paper → model → repo connections mined from source citations (Tier-1 exact match).
Why these links exist
Every edge carries a method, confidence, and the source snippet that justified it — so bad links are debuggable.
- PossiblePossibly related (embedding) · 56%GSRQ: Gain-Shape Residual Quantization for Sub-1-bit KV Cache →
- PossiblePossibly related (embedding) · 55%FreqDepthKV: Frequency-Guided Depth Sharing for Robust KV Cache Compression in Long-Context LLM Inference →
- PossiblePossibly related (embedding) · 52%BiSCo-LLM: Lookup-Free Binary Spherical Coding for Extreme Low-Bit Large Language Model Compression →
- PossiblePossibly related (embedding) · 50%DepthWeave-KV: Token-Adaptive Cross-Layer Residual Factorization for Long-Context KV Cache Compression →
- PossiblePossibly related (embedding) · 48%I mapped which local LLMs actually fit each RAM tier, 8 to 128GB (open dataset) →
- PossiblePossibly related (embedding) · 59%A JoLT for the KV Cache: Near-Lossless KV Cache Compression via Joint Tucker and JL-Residual Allocation for LLMs →
- PossiblePossibly related (embedding) · 47%Show HN: misa77 - a codec that decodes 2x faster than LZ4 (at better ratios) →
Implements
paperGSRQ: Gain-Shape Residual Quantization for Sub-1-bit KV CachepaperFreqDepthKV: Frequency-Guided Depth Sharing for Robust KV Cache Compression in Long-Context LLM InferencepaperBiSCo-LLM: Lookup-Free Binary Spherical Coding for Extreme Low-Bit Large Language Model CompressionpaperDepthWeave-KV: Token-Adaptive Cross-Layer Residual Factorization for Long-Context KV Cache Compression
Covers
Implements (incoming)
Covers (incoming)
Related across the graph
paperBiSCo-LLM: Lookup-Free Binary Spherical Coding for Extreme Low-Bit Large Language Model CompressionpaperA JoLT for the KV Cache: Near-Lossless KV Cache Compression via Joint Tucker and JL-Residual Allocation for LLMspaperDepthWeave-KV: Token-Adaptive Cross-Layer Residual Factorization for Long-Context KV Cache CompressionnewsI mapped which local LLMs actually fit each RAM tier, 8 to 128GB (open dataset)paperFreqDepthKV: Frequency-Guided Depth Sharing for Robust KV Cache Compression in Long-Context LLM InferencenewsShow HN: misa77 - a codec that decodes 2x faster than LZ4 (at better ratios)paperGSRQ: Gain-Shape Residual Quantization for Sub-1-bit KV Cache