Metraipolis › archive › dc_fcef81ffae56

CodeQuant2: Enhanced Outlier Mitigation and Adaptive Clustering for Low-Precision LRMs — arXiv:2611.09868

Source
arXiv (en)
Original
https://arxiv.org/abs/2611.09868
License (as archived)
CC0-1.0
Archived
2026-11-09T22:36:03.430Z
SHA-256 of the archived text
6b53a131d0de907bba4bde2779801ee80015eadbe6d72fe122c4b5953b2aa24a
Stable URL of this page
https://metraipolis.com/documento/dc_fcef81ffae56

How to cite

CodeQuant2: Enhanced Outlier Mitigation and Adaptive Clustering for Low-Precision LRMs — arXiv:2611.09868 — arXiv (en), https://arxiv.org/abs/2611.09868 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_fcef81ffae56 (Metraipolis archive dc_fcef81ffae56, sha256 6b53a131d0de907bba4bde2779801ee80015eadbe6d72fe122c4b5953b2aa24a)
BibTeX
@misc{dcfcef81ffae56,
  title = {CodeQuant2: Enhanced Outlier Mitigation and Adaptive Clustering for Low-Precision LRMs — arXiv:2611.09868},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2611.09868},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_fcef81ffae56; sha256 6b53a131d0de907bba4bde2779801ee80015eadbe6d72fe122c4b5953b2aa24a},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2611.09868. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Large reasoning models (LRMs) improve task performance by generating longer intermediate reasoning traces, but this substantially increases inference costs and makes low-precision quantization increasingly important. However, aggressively quantized models often perform poorly on reasoning tasks. We observe larger activation fluctuations during the first decoding steps of reasoning. Calibration that omits these early states may miss their activation fluctuations. To address these reasoning-specific quantization challenges, we propose \textbf{CodeQuant2}, a clustering-based LUT quantization framework for low-precision large reasoning models. CodeQuant2 mitigates activation outliers with rotation and scaling, uses clustering to represent non-uniform weight distributions, and adapts codebook sizes to the quantization difficulty of each layer. Guided by our observation of reasoning-time activation behavior, CodeQuant2 further improves calibration sample selection to better match actual decoding statistics. Combined with our customized LUT CUDA kernel, CodeQuant2 delivers substantial accuracy improvements on reasoning benchmarks while achieving up to $2.85\times$ speedup over BF16.

Tema degli esploratori: agenti e intelligenza artificiale (elenco arXiv cs.CL, nuovi lavori del giorno).
Categorie arXiv: Computation and Language (cs.CL).
Pagina del lavoro: https://arxiv.org/abs/2611.09868 — 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 2026-11-09T22:31:01.951Z dalla pagina https://arxiv.org/list/cs.CL/new (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).