Post-training Quantization for Hybrid Iterative Generative Models

arXiv 2608.13932 ↗First posted 14 Aug 2026cs.LGcited by 0

On the map: Synthetic data and self-improvement for LLMs → Model collapse from synthetic data

Abstract (arXiv)

Iterative Generative Models (IGMs) span autoregressive and diffusion paradigms, and hybrid variants that couple them can achieve remarkable image-generation fidelity. However, their iterative inference incurs substantial computational overhead, making Post-training Quantization (PTQ) appealing for acceleration, while directly applying vanilla PTQ to hybrid IGMs can trigger model collapse. By analyzing these failures, we identify two critical challenges: Excessive Outliers (EOs) in the activations create an irreconcilable trade-off between preserving normal precision and covering EOs, resulting in severe degradation in generation quality; Amplified Anomalies (AAs) arising unpredictably from minor quantization errors, create a mismatch between calibration and inference, thus iteratively triggering model collapse. To address these challenges, we introduce HyGenQ, a PTQ framework for hybrid IGMs. HyGenQ comprises Hierarchical Cluster Decoupling (HCD) and Scaling Recalibration (SR). HCD identifies and decouples outlier channels via a multi-stage clustering process, effectively isolating EOs while maintaining normal value precision, thereby alleviating performance degradation. SR scales AAs beyond Gaussian Bound, thereby avoiding model collapse caused by aggressive truncation. Extensive experiments demonstrate that HyGenQ successfully quantizes representative hybrid IGMs to 8-bit precision (W8A8), significantly outperforming existing baselines and validating its robustness across different model families.

Key facts. No citing papers in the map yet (data through 14 Oct 2026). Placed in Model collapse from synthetic data, one of 13 directions in Synthetic data and self-improvement for LLMs. Builds on 49 of its 76 references that are in the map. First posted on arXiv 14 Aug 2026.

Builds on

49 of 76 references.

References · 49

2017Attention Is All You Needcited by 60,361
2020An Image is Worth 16x16 Words: Transformers for Image Recognition at Scalecited by 30,186
2020Denoising Diffusion Probabilistic Modelscited by 21,076
2019PyTorch: An Imperative Style, High-Performance Deep Learning Librarycited by 20,305
2021High-Resolution Image Synthesis with Latent Diffusion Modelscited by 18,863
2021Diffusion Models Beat GANs on Image Synthesiscited by 8,255
2021Masked Autoencoders Are Scalable Vision Learnerscited by 7,403
2023Adding Conditional Control to Text-to-Image Diffusion Modelscited by 5,503
2021Improved Denoising Diffusion Probabilistic Modelscited by 3,761
2020Taming Transformers for High-Resolution Image Synthesiscited by 3,253
2016Improved Techniques for Training GANscited by 2,970
2022SmoothQuant: Accurate and Efficient Post-Training Quantization for Large Language Modelscited by 1,386

37 more

2022MaskGIT: Masked Generative Image Transformercited by 1,099
2016Pixel Recurrent Neural Networkscited by 816
2024Autoregressive Model Beats Diffusion: Llama for Scalable Image Generationcited by 777
2024Autoregressive Image Generation without Vector Quantizationcited by 651
2021ILVR: Conditioning Method for Denoising Diffusion Probabilistic Modelscited by 635
2024QuaRot: Outlier-Free 4-Bit Inference in Rotated LLMscited by 634
2019Learned Step Size Quantizationcited by 628
2020Up or Down? Adaptive Rounding for Post-Training Quantizationcited by 621
2021A White Paper on Neural Network Quantizationcited by 472
2019Q-BERT: Hessian Based Ultra Low Precision Quantization of BERTcited by 359
2023Q-Diffusion: Quantizing Diffusion Modelscited by 273
2022Post-training Quantization on Diffusion Modelscited by 267
2022QDrop: Randomly Dropping Quantization for Extremely Low-bit Post-Training Quantizationcited by 180
2021PTQ4ViT: Post-training quantization for vision transformers with twin uniform quantizationcited by 179
2023PTQD: Accurate Post-Training Quantization for Diffusion Modelscited by 170
2019Improving Neural Network Quantization without Retraining using Outlier Channel Splittingcited by 151
2017PixelSNAIL: An Improved Autoregressive Generative Modelcited by 137
2022RepQ-ViT: Scale Reparameterization for Post-Training Quantization of Vision Transformerscited by 137
2024HART: Efficient Visual Generation with Hybrid Autoregressive Transformercited by 132
2023RPTQ: Reorder-based Post-training Quantization for Large Language Modelscited by 116
2024ViDiT-Q: Efficient and Accurate Quantization of Diffusion Transformers for Image and Video Generationcited by 115
2022Q-ViT: Accurate and Fully Quantized Low-bit Vision Transformercited by 108
2023Temporal Dynamic Quantization for Diffusion Modelscited by 81
2024PTQ4DiT: Post-training Quantization for Diffusion Transformerscited by 79
2020Improved Techniques for Training Single-Image GANscited by 78
2022PD-Quant: Post-Training Quantization based on Prediction Difference Metriccited by 73
2023TFMQ-DM: Temporal Feature Maintenance Quantization for Diffusion Modelscited by 69
2013Deep AutoRegressive Networkscited by 61
2023Towards Accurate Post-training Quantization for Diffusion Modelscited by 47
2023On Error Propagation of Diffusion Modelscited by 31
2024TMPQ-DM: Joint Timestep Reduction and Quantization Precision Selection for Efficient Diffusion Modelscited by 10
2025Meaningless Tokens, Meaningful Gains: How Activation Shifts Enhance LLM Reasoningcited by 8
2025HadaNorm: Diffusion Transformer Quantization through Mean-Centered Transformationscited by 7
2024PQD: Post-training Quantization for Efficient Diffusion Modelscited by 6
2025Error Propagation Mechanisms and Compensation Strategies for Quantized Diffusioncited by 5
2024DiffLoss: unleashing diffusion model as constraint for training image restoration networkcited by 2
2024TaQ-DiT: Time-aware Quantization for Diffusion Transformerscited by 2

Built on by

Not cited yet.

Similar, not cited

Close in wording, no citation either way