Metraipolis › archive › dc_672109dd9bea

QuantWM: Temporally Consistent 2-Bit KV Cache Quantization for Video World Models — arXiv:2609.26425

Source
arXiv (en)
Original
https://arxiv.org/abs/2609.26425
License (as archived)
CC0-1.0
Archived
2026-10-08T13:47:12.521Z
SHA-256 of the archived text
52207dd341b1d07f3e72cd1e321507b337d4389e039409bf3aee6e076945fe4f
Stable URL of this page
https://metraipolis.com/documento/dc_672109dd9bea

How to cite

QuantWM: Temporally Consistent 2-Bit KV Cache Quantization for Video World Models — arXiv:2609.26425 — arXiv (en), https://arxiv.org/abs/2609.26425 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_672109dd9bea (Metraipolis archive dc_672109dd9bea, sha256 52207dd341b1d07f3e72cd1e321507b337d4389e039409bf3aee6e076945fe4f)
BibTeX
@misc{dc672109dd9bea,
  title = {QuantWM: Temporally Consistent 2-Bit KV Cache Quantization for Video World Models — arXiv:2609.26425},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2609.26425},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_672109dd9bea; sha256 52207dd341b1d07f3e72cd1e321507b337d4389e039409bf3aee6e076945fe4f},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.26425. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Video world models achieve long-range temporal consistency by storing KV cache during generation, but the growing cache makes KV cache memory a major deployment bottleneck, which motivates low-bit quantization study for efficiency. Existing 2-bit KV cache quantization methods can achieve nearly lossless performance on VBench, however, when applied to video world models, we find they still cause severe temporal flickering and visual degradation. Meanwhile, deeper investigates show that Key quantization produces smaller reconstruction errors than Value, but surprisingly leads to larger output degradation. We trace this discrepancy to attention in video world models: Key perturbations can change the attention logits, and shift the temporal-spatial tokens selected by Queries. These observations motivate us to preserve attention logits and temporal-spatial token selection during KV cache quantization. To address this issue, we present QuantWM, a training-free 2-bit KV cache quantization framework for video world models. QuantWM introduces two complementary techniques to mitigate the attention shifts. Firstly, quantization-sensitivity-aware clustering (QSAC) jointly considers historical Query sensitivity and residual ranges to select INT2-friendly Key centroids, which reduces quantization errors in channels that are more critical to attention. In addition, principal-subspace attention compensation (PSAC) restores the remaining Key errors along the dominant Query subspace using low-rank projections, which provides a direct and efficient correction to stabilize attention logits. Experiments on LingBot-World-v2, HY-World 1.5, Matrix-Game-2, Longcat-Video and Causal-Forcing demonstrate that QuantWM significantly improves visual quality and temporal consistency, while outperforming existing methods across benchmarks with up to 6.20 KV cache memory compression and limited additional overhead.

Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-09-23 in cs.AI).
Categorie arXiv: Computer Vision and Pattern Recognition (cs.CV) ; Artificial Intelligence (cs.AI).
Pagina del lavoro: https://arxiv.org/abs/2609.26425 — 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 2026-10-08T13:18:38.314Z dalla pagina https://arxiv.org/catchup/cs.AI/2026-09-23?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).