{
  "annotation_id": "44a126e8-1d79-4652-b2b3-9df4d63fc06a",
  "date_created": "2026-02-17T05:53:08.155000Z",
  "date_modified": "2026-02-17T05:53:08.155000Z",
  "file_hash": "7f66606479955092e5f3ec17a3612f35d24ca19fc58a205a2ba8ee28bdb48a04",
  "private": false,
  "record": {
    "abstract": "Post-training quantization is essential for deploying Large Language Models (LLMs) on resource-constrained devices. However, standard integer quantization (e.g., INT4) fundamentally degrades performance by imposing a uniform grid on the heavy-tailed distribution of weight parameters, particularly in smaller-scale models (e.g., \u003c2B parameters). We introduce HAS-VQ (Hessian-Adaptive Sparse Vector Quantization), a compression framework that strictly decouples high-sensitivity outliers from the bulk weight distribution using second-order sensitivity analysis. HAS-VQ employs a Hessian-Masked Decoupling strategy to isolate sensitive parameters, followed by robust Vector Quantization (VQ) of the remaining dense body. Crucially, we introduce a residual sparse feedback mechanism that corrects quantization errors in the most sensitive dimensions, ensuring exact reconstruction of outliers. We evaluate HAS-VQ on SmolLM2-1.7B, demonstrating two distinct regimes of superiority: (1) Pareto Dominance over Integer Baselines: At 4.23 effective bits-per-parameter (BPP), we achieve a perplexity of 14.23, significantly outperforming the standard INT4 baseline (20.03 PPL at 4.71 BPP). (2) High-Fidelity Compression: Relative to the FP16 baseline, HAS-VQ achieves a 2.3x reduction in model size (7.03 BPP) while maintaining statistically indistinguishable perplexity (10.12 vs. 10.04), effectively offering a lossless compression alternative for bandwidth-constrained environments. The code is available at https://github.com/VladimerKhasia/HASVQ",
    "arxiv_id": "2601.06959",
    "authors": [
      "Vladimer Khasia"
    ],
    "categories": [
      "cs.LG"
    ],
    "license": "http://arxiv.org/licenses/nonexclusive-distrib/1.0/",
    "title": "HAS-VQ: Hessian-Adaptive Sparse Vector Quantization for High-Fidelity LLM Compression",
    "url": "https://arxiv.org/abs/2601.06959",
    "version": "v1"
  },
  "schema_id": "dorsal/arxiv",
  "source": {
    "execution_id": "ae802eef-fe20-4ff9-a080-c5e42690cbc8",
    "id": "arXiv Dataset",
    "type": "Model",
    "variant": "snapshot-2026-01-17",
    "version": "0.1.0"
  },
  "user_id": 1000002
}