Metraipolis › archive › dc_a5d033c5288b
PrismQuant: Optimal Null-Space Rotations for Grouped Quantizers — arXiv:2609.32429
- Source
- arXiv (en)
- Original
- https://arxiv.org/abs/2609.32429
- License (as archived)
- CC0-1.0
- Archived
- 2026-10-07T20:55:30.003Z
- SHA-256 of the archived text
32f82157e8fe96c471280417debceaecbef8bec131fa71fdee0c2be4aacd7333- Stable URL of this page
https://metraipolis.com/documento/dc_a5d033c5288b
How to cite
PrismQuant: Optimal Null-Space Rotations for Grouped Quantizers — arXiv:2609.32429 — arXiv (en), https://arxiv.org/abs/2609.32429 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_a5d033c5288b (Metraipolis archive dc_a5d033c5288b, sha256 32f82157e8fe96c471280417debceaecbef8bec131fa71fdee0c2be4aacd7333)
BibTeX
@misc{dca5d033c5288b,
title = {PrismQuant: Optimal Null-Space Rotations for Grouped Quantizers — arXiv:2609.32429},
howpublished = {arXiv (en)},
url = {https://arxiv.org/abs/2609.32429},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_a5d033c5288b; sha256 32f82157e8fe96c471280417debceaecbef8bec131fa71fdee0c2be4aacd7333},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.32429. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
Smaller activation outliers do not necessarily imply better low-bit quantization: their alignment with the quantizer matters. We introduce PrismQuant, a quantizer-aware rotation framework that aligns the leading activation eigenspace with the constant group subspace of asymmetric grouped INT4. The affine offsets represent the energy in this subspace without widening the range within the group. We formulate rotation design as a Ky Fan trace maximization and derive a closed-form solution that is provably optimal for this alignment objective. Compact Householder transformations and their compact-WY representation enable gradient-free construction and efficient application at both foldable and online sites. A predictive range law further connects unaligned activation energy and group size to quantization-relevant variation. Experiments on Llama, Qwen, and Mistral span dense models up to 70B parameters and a 30B mixture-of-experts model. Under W4A4KV4, PrismQuant sets the state of the art on Llama-3.2-3B among the compared methods in both perplexity and accuracy. On Llama-3.1-70B, it attains 3.85 perplexity and 72.46% average zero-shot accuracy, only 0.22 percentage points below full precision. In the deployment study on Llama-3.1-8B, our optimized implementation achieves 1.51x prefill and 1.22x CUDA Graph decode speedups over matched FP16 baselines, with 56.34% lower decode peak memory and only 2.35% additional Graph decode latency over Hadamard. Code is available at this https URL . Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-09-29 in cs.AI). Categorie arXiv: Artificial Intelligence (cs.AI) ; Machine Learning (cs.LG). Pagina del lavoro: https://arxiv.org/abs/2609.32429 — 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 2026-10-07T20:29:00.794Z dalla pagina https://arxiv.org/catchup/cs.AI/2026-09-29?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).