TASQ: Temporal-Adaptive Bit Sparsification Quantization for Diffusion Models
Quick Answer
TASQ introduces Temporal-Adaptive Bit Sparsification Quantization, optimizing denoising in models like PixArt-Sigma and SDXL-Turbo.
Quick Take
It reduces execution cycles by 25-50% compared to static quantization, achieving similar quality while lowering computational costs.
Key Points
- TASQ uses a shared maximum-precision weight buffer for efficient quantization.
- It learns a Temporal-Spatial LSB Mask to adapt precision per denoising stage.
- Execution cycles are reduced by 6.1 to 7.5x over naive static 8-bit execution.
- Quality remains comparable to static quantization despite lower computation.
- Code for TASQ is publicly available for further research.
Paper Resources
📖 Reader Mode
~2 min readAbstract:Static quantization assigns one weight precision to every denoising step. To preserve quality, that precision must accommodate the most quantization-sensitive step, even though many other steps can tolerate fewer bits. The resulting model may satisfy its memory budget, but it repeatedly pays worst-case arithmetic throughout the denoising trajectory. We introduce Temporal-Adaptive Bit Sparsification Quantization (TASQ) to separate these two costs. TASQ stores one shared maximum-precision weight buffer and learns a Temporal-Spatial LSB Mask that selects a lower effective precision for each layer and denoising stage by truncating least-significant bits. Storage therefore remains fixed by the worst case, while BitOPs decrease at less sensitive stages without per-stage weight copies or runtime search. A Temporal-Precision Engine maps the learned schedule to bit-serial execution, where cycles scale with effective precision and switching precision has no measured cycle overhead. On PixArt-Sigma, SANA-1.6B, and SDXL-Turbo, TASQ achieves quality comparable to static quantization with less computation. Together with the Temporal-Precision Engine, it reduces execution cycles by 25 to 50 percent over static quantization and by 6.1 to 7.5x over a naive static 8-bit bit-serial execution. Code is available at this https URL.
| Subjects: | Computer Vision and Pattern Recognition (cs.CV) |
| Cite as: | arXiv:2608.03057 [cs.CV] |
| (or arXiv:2608.03057v1 [cs.CV] for this version) | |
| https://doi.org/10.48550/arXiv.2608.03057 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Seokho Han [view email]
[v1]
Tue, 4 Aug 2026 03:14:48 UTC (7,570 KB)
— Originally published at arxiv.org
Want this in your inbox every morning?
Daily brief at your local 8am — bilingual EN/中文, free.
More from arXiv cs.CV
See more →ProMoE-FL: Prototype-conditioned Mixture of Experts for Multimodal Federated Learning with Missing Modalities
ProMoE-FL introduces a Prototype-conditioned Mixture-of-Experts framework for multimodal federated learning, effectively addressing missing modalities. It outperforms existing methods on four chest X-ray datasets, demonstrating superior feature synthesis capabilities in both homogeneous and heterogeneous settings.