Metraipolis › archive › dc_821d41dfd826
KV Cache Quantization for Diffusion Language Models — arXiv:2610.39853
- Source
- arXiv (en)
- Original
- https://arxiv.org/abs/2610.39853
- License (as archived)
- CC0-1.0
- Archived
- 2026-10-31T06:17:03.322Z
- SHA-256 of the archived text
e4c7ae516ec05d4627b89bb62d846d60eae285866c2642eb01b41b9e35dafaf8- Stable URL of this page
https://metraipolis.com/documento/dc_821d41dfd826
How to cite
KV Cache Quantization for Diffusion Language Models — arXiv:2610.39853 — arXiv (en), https://arxiv.org/abs/2610.39853 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_821d41dfd826 (Metraipolis archive dc_821d41dfd826, sha256 e4c7ae516ec05d4627b89bb62d846d60eae285866c2642eb01b41b9e35dafaf8)
BibTeX
@misc{dc821d41dfd826,
title = {KV Cache Quantization for Diffusion Language Models — arXiv:2610.39853},
howpublished = {arXiv (en)},
url = {https://arxiv.org/abs/2610.39853},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_821d41dfd826; sha256 e4c7ae516ec05d4627b89bb62d846d60eae285866c2642eb01b41b9e35dafaf8},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2610.39853. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
KV cache quantization is well studied for autoregressive language models, where cached prefix states are exact under causal attention and quantization is mainly a cache reconstruction problem. Diffusion language models (dLLMs) break this assumption. Fast dLLM inference reuses KV states across denoising steps, but bidirectional token representations can change as masked tokens are revealed, so the reused cache may already differ from full recomputation before quantization is applied. We study KV cache quantization under this cache-reuse error. We adapt KIVI and KVQuant to Fast-dLLM-style prefix and dual cache reuse, and introduce TRACE and dLLM-KVQ. TRACE reconstructs feasible dLLM reveal trajectories from prompt-answer pairs, providing calibration states that match iterative masked decoding. dLLM-KVQ uses these trajectories to choose per-layer K-cache entries that remain in high precision according to output- and attention-level behavior metrics, while applying standard low-bit value quantization. Across LLaDA and Dream evaluations on GSM8K, HumanEval, MATH, and MBPP, dLLM-KVQ achieves the highest average score among quantized methods across the reported 4-bit and 2-bit prefix/dual cache settings. 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/2610.39853 — 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 2026-10-31T06:07:23.364Z dalla pagina https://arxiv.org/list/cs.CL/new (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).