Metraipolis › archive › dc_6da479e87ec3
Different Tasks, Different Timesteps: Temporal Task-Aware Quantization for Diffusion Language Models — arXiv:2610.31198
- Source
- arXiv (en)
- Original
- https://arxiv.org/abs/2610.31198
- License (as archived)
- CC0-1.0
- Archived
- 2026-10-25T08:33:23.310Z
- SHA-256 of the archived text
9e43264cde12a00fc5868e7ca3128f5f201ac0c5641a89299b1030ca24128a8e- Stable URL of this page
https://metraipolis.com/documento/dc_6da479e87ec3
How to cite
Different Tasks, Different Timesteps: Temporal Task-Aware Quantization for Diffusion Language Models — arXiv:2610.31198 — arXiv (en), https://arxiv.org/abs/2610.31198 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_6da479e87ec3 (Metraipolis archive dc_6da479e87ec3, sha256 9e43264cde12a00fc5868e7ca3128f5f201ac0c5641a89299b1030ca24128a8e)
BibTeX
@misc{dc6da479e87ec3,
title = {Different Tasks, Different Timesteps: Temporal Task-Aware Quantization for Diffusion Language Models — arXiv:2610.31198},
howpublished = {arXiv (en)},
url = {https://arxiv.org/abs/2610.31198},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_6da479e87ec3; sha256 9e43264cde12a00fc5868e7ca3128f5f201ac0c5641a89299b1030ca24128a8e},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2610.31198. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
Diffusion Language Models (DLMs) introduce unique quantization challenges due to their iterative denoising process. In this work, we reveal that quantization sensitivity is both taskand time-dependent, with different downstream tasks exhibiting distinct precision requirements across denoising steps. We further show that these patterns can be predicted without groundtruth labels or task-specific evaluation metrics. Building on these findings, we introduce T 2Quant, a task and time-aware mixedprecision quantization framework that uses a label-free sensitivity signal to optimally allocate activation precision across denoising steps under a fixed bit-width budget. Extensive experiments across three DLMs and six benchmarks demonstrate that T 2Quant consistently improves the average accuracy of existing quantization methods, achieving gains of up to 6.7 percentage points under the same budget. Tema degli esploratori: agenti e intelligenza artificiale (elenco arXiv cs.AI, nuovi lavori del giorno). Categorie arXiv: Artificial Intelligence (cs.AI). Pagina del lavoro: https://arxiv.org/abs/2610.31198 — 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 2026-10-25T08:31:13.073Z dalla pagina https://arxiv.org/list/cs.AI/new (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).