Metraipolis › archive › dc_8dfbfc3773ff

JARQ: Joint Alternating Refinement for Quantization — arXiv:2609.38599

Source: arXiv (en)
Original: https://arxiv.org/abs/2609.38599
License (as archived): CC0-1.0
Archived: 2026-10-07T13:45:24.811Z
SHA-256 of the archived text: 039d1af57853b5f2a6c83161cfebf447e4b7e373395f3bb19152fb879ac28c64
Stable URL of this page: https://metraipolis.com/documento/dc_8dfbfc3773ff

How to cite

JARQ: Joint Alternating Refinement for Quantization — arXiv:2609.38599 — arXiv (en), https://arxiv.org/abs/2609.38599 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_8dfbfc3773ff (Metraipolis archive dc_8dfbfc3773ff, sha256 039d1af57853b5f2a6c83161cfebf447e4b7e373395f3bb19152fb879ac28c64)

BibTeX

@misc{dc8dfbfc3773ff,
  title = {JARQ: Joint Alternating Refinement for Quantization — arXiv:2609.38599},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2609.38599},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_8dfbfc3773ff; sha256 039d1af57853b5f2a6c83161cfebf447e4b7e373395f3bb19152fb879ac28c64},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.38599. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

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.

Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-10-01 in cs.LG).
Categorie arXiv: Machine Learning (cs.LG).
Pagina del lavoro: https://arxiv.org/abs/2609.38599 — autori e testo completo sono lì (qui omessi i nomi: nessun dato personale nel deposito).
Licenza: titolo e abstract sono metadati arXiv distribuiti in CC0 1.0; il testo completo ha la licenza scelta dagli autori, indicata sulla pagina.
Letto il 2026-10-07T13:18:10.232Z dalla pagina https://arxiv.org/catchup/cs.LG/2026-10-01?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).