# Qua2SeDiMo: Quantifiable Quantization Sensitivity of Diffusion Models

arXiv: https://arxiv.org/abs/2412.14628 · HTML: https://mlsanity.com/paper/2412.14628/

## Key facts

- Cited by 7 papers in the map (data through 14 Oct 2026).
- Placed in Diffusion foundations and sampling, one of 14 directions in Diffusion models.
- Builds on 11 of its 55 references that are in the map.
- First posted on arXiv 19 Dec 2024.

## Placed in

- Diffusion models / [Diffusion foundations and sampling](https://mlsanity.com/diffusion/diffusion-foundations-and-sampling/)

## Builds on (most cited first, 11 of 11)

- [Llama 2: Open Foundation and Fine-Tuned Chat Models](https://mlsanity.com/paper/2307.09288/) (2023)
- [Auto-Encoding Variational Bayes](https://mlsanity.com/paper/1312.6114/) (2013)
- [CLIPScore: A Reference-free Evaluation Metric for Image Captioning](https://mlsanity.com/paper/2104.08718/) (2021)
- [Quantizing deep convolutional networks for efficient inference: A whitepaper](https://mlsanity.com/paper/1806.08342/) (2018)
- [The Era of 1-bit LLMs: All Large Language Models are in 1.58 Bits](https://mlsanity.com/paper/2402.17764/) (2024)
- [Hunyuan-DiT: A Powerful Multi-Resolution Diffusion Transformer with Fine-Grained Chinese Understanding](https://mlsanity.com/paper/2405.08748/) (2024)
- [The Unreasonable Ineffectiveness of the Deeper Layers](https://mlsanity.com/paper/2403.17887/) (2024)
- [ViDiT-Q: Efficient and Accurate Quantization of Diffusion Transformers for Image and Video Generation](https://mlsanity.com/paper/2406.02540/) (2024)
- [QuEST: Low-bit Diffusion Model Quantization via Efficient Selective Finetuning](https://mlsanity.com/paper/2402.03666/) (2024)
- [Model Quantization and Hardware Acceleration for Vision Transformers: A Comprehensive Survey](https://mlsanity.com/paper/2405.00314/) (2024)
- [FRAP: Faithful and Realistic Text-to-Image Generation with Adaptive Prompt Weighting](https://mlsanity.com/paper/2408.11706/) (2024)

## Built on by (most cited first, 7 of 7)

- [Low-bit Model Quantization for Deep Neural Networks: A Survey](https://mlsanity.com/paper/2505.05530/) (2025)
- [Re-ttention: Ultra Sparse Visual Generation via Attention Statistical Reshape](https://mlsanity.com/paper/2505.22918/) (2025)
- [FP4DiT: Towards Effective Floating Point Quantization for Diffusion Transformers](https://mlsanity.com/paper/2503.15465/) (2025)
- [TreeQ: Pushing the Quantization Boundary of Diffusion Transformer via Tree-Structured Mixed-Precision Search](https://mlsanity.com/paper/2512.06353/) (2025)
- [Na"ive PAINE: Lightweight Text-to-Image Generation Improvement with Prompt Evaluation](https://mlsanity.com/paper/2603.12506/) (2026)
- [Can We Predict The Human Preference For Text-to-Image Content Prior To Generation And Is It Even Useful To Do So?](https://mlsanity.com/paper/2606.05478/) (2026)
- [Layerwise Error Attribution for Fast and Robust Mixed-Precision Post-Training Quantization](https://mlsanity.com/paper/2610.09877/) (2026)

## Abstract (arXiv)

Diffusion Models (DM) have democratized AI image generation through an iterative denoising process. Quantization is a major technique to alleviate the inference cost and reduce the size of DM denoiser networks. However, as denoisers evolve from variants of convolutional U-Nets toward newer Transformer architectures, it is of growing importance to understand the quantization sensitivity of different weight layers, operations and architecture types to performance. In this work, we address this challenge with Qua2SeDiMo, a mixed-precision Post-Training Quantization framework that generates explainable insights on the cost-effectiveness of various model weight quantization methods for different denoiser operation types and block structures. We leverage these insights to make high-quality mixed-precision quantization decisions for a myriad of diffusion models ranging from foundational U-Nets to state-of-the-art Transformers. As a result, Qua2SeDiMo can construct 3.4-bit, 3.9-bit, 3.65-bit and 3.7-bit weight quantization on PixArt-α, PixArt-Sigma, Hunyuan-DiT and SDXL, respectively. We further pair our weight-quantization configurations with 6-bit activation quantization and outperform existing approaches in terms of quantitative metrics and generative image quality.