analiza aiqbench
JARQ introduces a novel post-training quantization refinement technique that jointly optimizes group scales and integer codes without requiring backpropagation, addressing accuracy losses inherent in standard grid-based rounding methods. By effectively coupling error corrections across groups, the method significantly reduces perplexity and improves multiple-choice accuracy across major LLM architectures like Llama-3 and Qwen while maintaining inference costs. This advancement offers a practical pathway to deploying smaller-bit quantized models with performance closer to full-precision counterparts.
JARQ: Joint Alternating Refinement for Quantization
arXiv:2609.38599v1 Announce Type: new Abstract: Group-wise post-training quantizers for large language models round weights onto a grid that is not refit to the resulting integer codes. We show that this leaves accuracy on the table: the best grid depends on the codes, input correlations couple the errors of different groups, and useful code changes often involve many codes at once. We propose JARQ , a plug-in refinement that starts from any group-wise quantizer and alternates a joint least-squares fit of all group scales with bounded Babai proposals that move many codes of a group together on the current grid. The problem is a bilinear box-constrained mixed-integer least-squares problem; the solver is backpropagation-free, does not increase the layer-wise objective under exact scale solves, and keeps the host's bit width, groups, zero points, and inference cost. Across Llama-2, Llama-3, and Qwen models with RTN, GPTQ, OmniQuant, and AWQ hosts, JARQ lowers perplexity in 90 of 96 comparisons, cuts three-bit RTN perplexity by up to 36%, raises mean multiple-choice accuracy in 23 of 24 configurations, and improves QEP, QuaRot, and OJBKQ outputs, at under a minute per 7B block.
Abstract:Group-wise post-training quantizers for large language models round weights onto a grid that is not refit to the resulting integer codes. We show that this leaves accuracy on the table: the best grid depends on the codes, input correlations couple the errors of different groups, and useful code changes often involve many codes at once. We propose JARQ , a plug-in refinement that starts from any group-wise quantizer and alternates a joint least-squares fit of all group scales with bounded Babai proposals that move many codes of a group together on the current grid. The problem is a bilinear box-constrained mixed-integer least-squares problem; the solver is backpropagation-free, does not increase the layer-wise objective under exact scale solves, and keeps the host's bit width, groups, zero points, and inference cost. Across Llama-2, Llama-3, and Qwen models with RTN, GPTQ, OmniQuant, and AWQ hosts, JARQ lowers perplexity in 90 of 96 comparisons, cuts three-bit RTN perplexity by up to 36%, raises mean multiple-choice accuracy in 23 of 24 configurations, and improves QEP, QuaRot, and OJBKQ outputs, at under a minute per 7B block.
Machine Learning (cs.LG)
arXiv:2609.38599 [cs.LG]
(or arXiv:2609.38599v1 [cs.LG] for this version)
https://doi.org/10.48550/arXiv.2609.38599 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Text integral sub licența deschisă a publicației · 275 cuvinte
Subiecte conexe
[AINews] Reality Checks on AI News (Yegge shuts down Gas Town, Databricks’ +60% Astra cost)
Latent Space · 17 sept. 2026
FACT: A Multimodal Force-Action-Track Foundation Model for Human-to-Robot Dexterity
arXiv · 15 oct. 2026
BARQ: Balanced Codebook Refinement for Low-Bit LLM Quantization
arXiv · 6 oct. 2026
LLM Compression with Jointly Optimizing Architectural and Quantization choices
arXiv · 5 iun. 2026