Q-Sched: Pushing the Boundaries of Few-Step Diffusion Models with Quantization-Aware Scheduling

Natalia Frumkin, Diana Marculescu
Proceedings of the 43rd International Conference on Machine Learning, PMLR 306:31703-31725, 2026.

Abstract

Text-to-image diffusion models remain computationally intensive: generating a single image typically requires dozens of passes through large transformer backbones (e.g., SDXL uses 50 evaluations of a 2.6B-parameter model). Few-step variants reduce the step count to 2–8 but still rely on large, full-precision backbones, making inference impractical on resource-constrained platforms. Existing post-training quantization (PTQ) methods are further hampered by their dependence on full-precision calibration. We introduce Q-Sched, a scheduler-level PTQ approach that adapts the diffusion sampler while keeping the quantized weights fixed. By adjusting the few-step sampling trajectory with quantization-aware preconditioning coefficients, Q-Sched matches or surpasses full-precision quality while delivering a 4$\times$ reduction in model size and preserving a single reusable checkpoint across bit-widths. To learn these coefficients, we propose a reference-free Joint Alignment–Quality (JAQ) loss, which combines text–image compatibility with an image-quality objective for fine-grained control. JAQ requires only a handful of calibration prompts and avoids any full-precision inference during calibration. Empirically, Q-Sched yields substantial gains: a 15.5% FID improvement over the FP16 4-step Latent Consistency Model and a 16.6% improvement over the FP16 8-step Phased Consistency Model, demonstrating that quantization and few-step distillation are complementary for high-fidelity generation. A large-scale user study with 80,000+ annotations further validates these results on both FLUX.1[schnell] and SDXL-Turbo. Code: https://github.com/enyac-group/q-sched

Cite this Paper

BibTeX
@InProceedings{pmlr-v306-frumkin26a, title = {Q-Sched: Pushing the Boundaries of Few-Step Diffusion Models with Quantization-Aware Scheduling}, author = {Frumkin, Natalia and Marculescu, Diana}, booktitle = {Proceedings of the 43rd International Conference on Machine Learning}, pages = {31703--31725}, year = {2026}, editor = {Zhang, Tong and Dudik, Miroslav and Jaggi, Martin and Agarwal, Alekh and Li, Sharon and Schuurmans, Dale and Zhu, Jerry and Berkenkamp, Felix and Dong, Hanze and Bietti, Alberto}, volume = {306}, series = {Proceedings of Machine Learning Research}, month = {06--11 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v306/main/assets/frumkin26a/frumkin26a.pdf}, url = {https://proceedings.mlr.press/v306/frumkin26a.html}, abstract = {Text-to-image diffusion models remain computationally intensive: generating a single image typically requires dozens of passes through large transformer backbones (e.g., SDXL uses 50 evaluations of a 2.6B-parameter model). Few-step variants reduce the step count to 2–8 but still rely on large, full-precision backbones, making inference impractical on resource-constrained platforms. Existing post-training quantization (PTQ) methods are further hampered by their dependence on full-precision calibration. We introduce Q-Sched, a scheduler-level PTQ approach that adapts the diffusion sampler while keeping the quantized weights fixed. By adjusting the few-step sampling trajectory with quantization-aware preconditioning coefficients, Q-Sched matches or surpasses full-precision quality while delivering a 4$\times$ reduction in model size and preserving a single reusable checkpoint across bit-widths. To learn these coefficients, we propose a reference-free Joint Alignment–Quality (JAQ) loss, which combines text–image compatibility with an image-quality objective for fine-grained control. JAQ requires only a handful of calibration prompts and avoids any full-precision inference during calibration. Empirically, Q-Sched yields substantial gains: a 15.5% FID improvement over the FP16 4-step Latent Consistency Model and a 16.6% improvement over the FP16 8-step Phased Consistency Model, demonstrating that quantization and few-step distillation are complementary for high-fidelity generation. A large-scale user study with 80,000+ annotations further validates these results on both FLUX.1[schnell] and SDXL-Turbo. Code: https://github.com/enyac-group/q-sched} }
Endnote
%0 Conference Paper %T Q-Sched: Pushing the Boundaries of Few-Step Diffusion Models with Quantization-Aware Scheduling %A Natalia Frumkin %A Diana Marculescu %B Proceedings of the 43rd International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2026 %E Tong Zhang %E Miroslav Dudik %E Martin Jaggi %E Alekh Agarwal %E Sharon Li %E Dale Schuurmans %E Jerry Zhu %E Felix Berkenkamp %E Hanze Dong %E Alberto Bietti %F pmlr-v306-frumkin26a %I PMLR %P 31703--31725 %U https://proceedings.mlr.press/v306/frumkin26a.html %V 306 %X Text-to-image diffusion models remain computationally intensive: generating a single image typically requires dozens of passes through large transformer backbones (e.g., SDXL uses 50 evaluations of a 2.6B-parameter model). Few-step variants reduce the step count to 2–8 but still rely on large, full-precision backbones, making inference impractical on resource-constrained platforms. Existing post-training quantization (PTQ) methods are further hampered by their dependence on full-precision calibration. We introduce Q-Sched, a scheduler-level PTQ approach that adapts the diffusion sampler while keeping the quantized weights fixed. By adjusting the few-step sampling trajectory with quantization-aware preconditioning coefficients, Q-Sched matches or surpasses full-precision quality while delivering a 4$\times$ reduction in model size and preserving a single reusable checkpoint across bit-widths. To learn these coefficients, we propose a reference-free Joint Alignment–Quality (JAQ) loss, which combines text–image compatibility with an image-quality objective for fine-grained control. JAQ requires only a handful of calibration prompts and avoids any full-precision inference during calibration. Empirically, Q-Sched yields substantial gains: a 15.5% FID improvement over the FP16 4-step Latent Consistency Model and a 16.6% improvement over the FP16 8-step Phased Consistency Model, demonstrating that quantization and few-step distillation are complementary for high-fidelity generation. A large-scale user study with 80,000+ annotations further validates these results on both FLUX.1[schnell] and SDXL-Turbo. Code: https://github.com/enyac-group/q-sched
APA
Frumkin, N. & Marculescu, D.. (2026). Q-Sched: Pushing the Boundaries of Few-Step Diffusion Models with Quantization-Aware Scheduling. Proceedings of the 43rd International Conference on Machine Learning, in Proceedings of Machine Learning Research 306:31703-31725 Available from https://proceedings.mlr.press/v306/frumkin26a.html.

Related Material