mlsanityData to 7 Oct 2026

Chameleon: Dynamic Format Adapter for Efficient Diffusion

arXiv 2609.33496 ↗First posted 27 Sep 2026cs.LG cs.AI cs.CVcited by 0

On the map: Diffusion models → Diffusion foundations and sampling

Abstract (arXiv)

Post-training quantization (PTQ) is the standard way to run modern diffusion models on memory-constrained accelerators, yet every existing diffusion PTQ scheme fixes the mathitnumber format in advance and only tunes the scale, zero point, or per-layer bit-width. At a fixed bit-width the best format depends on the distribution being encoded, and that distribution differs across weight channels, across layers, and along the diffusion timestep, where activation distributions slide from heavy-tailed and noise-dominated to tightly clustered and structured. We propose Chameleon, a PTQ framework that holds the bit-width fixed and treats the format itself as a discrete variable, chosen per weight channel and per (layer, timestep bucket) activation tensor. Activation formats come from INT8, FP8 E4M3, FP8 E5M2, MXFP8, MXINT8, selected ahead of time from two cheap statistics (empirical kurtosis and the closed-form diffusion SNR) and stored in a lookup table; weight formats come from INT8, MXINT8 at 8 bits or INT4, NF4, FP4 E2M1, MXINT4, MXFP4 at 4 bits, selected offline by reconstruction error. An architectural fork adapts the same selection layer to multi-step UNets, single-step distilled models, and Diffusion Transformers. Across SDXL, SDXL-Turbo, and PixArt-α on COCO-2014, Chameleon achieves the best FID in all six backbone × bit-width settings, with CLIP within 0.24 of the FP16 reference and the best of all quantized methods at W4A8.

Key facts. No citing papers in the map yet (data through 7 Oct 2026). Placed in Diffusion foundations and sampling, one of 14 directions in Diffusion models. Builds on 28 of its 34 references that are in the map. First posted on arXiv 27 Sep 2026.

Builds on

28 of 34 references.

Closest ancestors · 17

Same direction, newest first

  1. 2025Q-Sched: Pushing the Boundaries of Few-Step Diffusion Models with Quantization-Aware Scheduling1 Sep
  2. 2025FP4DiT: Towards Effective Floating Point Quantization for Diffusion Transformers19 Mar
  3. 2024Q-DiT: Accurate Post-Training Quantization for Diffusion Transformers25 Jun
  4. 2024ViDiT-Q: Efficient and Accurate Quantization of Diffusion Transformers for Image and Video Generation4 Jun
  5. 2024MixDQ: Memory-Efficient Few-Step Text-to-Image Diffusion Models with Metric-Decoupled Mixed Precision Quantization28 May
  6. 2024PTQ4DiT: Post-training Quantization for Diffusion Transformers25 May
  7. 2024BinaryDM: Accurate Weight Binarization for Efficient Diffusion Models8 Apr
  8. 2023Adversarial Diffusion Distillation28 Nov
  9. 2023Latent Consistency Models: Synthesizing High-Resolution Images with Few-Step Inference6 Oct
  10. 2023EfficientDM: Efficient Quantization-Aware Fine-Tuning of Low-Bit Diffusion Models5 Oct
  11. 2023PixArt-α: Fast Training of Diffusion Transformer for Photorealistic Text-to-Image Synthesis30 Sep
  12. 2023SDXL: Improving Latent Diffusion Models for High-Resolution Image Synthesis4 Jul
5 more
  1. 2023Temporal Dynamic Quantization for Diffusion Models4 Jun
  2. 2023PTQD: Accurate Post-Training Quantization for Diffusion Models18 May
  3. 2023Q-Diffusion: Quantizing Diffusion Models8 Feb
  4. 2022Post-training Quantization on Diffusion Models28 Nov
  5. 2022DPM-Solver: A Fast ODE Solver for Diffusion Probabilistic Model Sampling in Around 10 Steps2 Jun

Other references · 11

  1. 2021Learning Transferable Visual Models From Natural Language Supervisioncited by 33,456
  2. 2014Microsoft COCO: Common Objects in Contextcited by 17,498
  3. 2022SmoothQuant: Accurate and Efficient Post-Training Quantization for Large Language Modelscited by 1,338
  4. 2023AWQ: Activation-aware Weight Quantization for LLM Compression and Accelerationcited by 1,234
  5. 2021A White Paper on Neural Network Quantizationcited by 453
  6. 2022FP8 Quantization: The Power of the Exponentcited by 87
  7. 2018Rethinking floating point for deep learningcited by 22
  8. 2023Integer or Floating Point? New Outlooks for Low-Bit Quantization on Large Language Modelscited by 18
  9. 2019Neural Network Training with Approximate Logarithmic Computationscited by 7
  10. 2025OASIS: Optimized Lightweight Autoencoder System for Distributed In-Sensor computingcited by 2
  11. 2025EntroLLM: Entropy Encoded Weight Compression for Efficient Large Language Model Inference on Edge Devicescited by 1

Built on by

Not cited yet.

Similar, not cited

Close in wording, no citation either way

  1. 2025DiffPro: Joint Timestep and Layer-Wise Precision Optimization for Efficient Diffusion Inference
  2. 2025SQ-DM: Accelerating Diffusion Models with Aggressive Quantization and Temporal Sparsity
  3. 2024Efficiency Meets Fidelity: A Novel Quantization Framework for Stable Diffusion
  4. 2024TCAQ-DM: Timestep-Channel Adaptive Quantization for Diffusion Models
  5. 2026TASQ: Temporal-Adaptive Bit Sparsification Quantization for Diffusion Models
  6. 2024TMPQ-DM: Joint Timestep Reduction and Quantization Precision Selection for Efficient Diffusion Models
  7. 2024PQD: Post-training Quantization for Efficient Diffusion Models
  8. 2026TMPDiff: Temporal Mixed-Precision for Diffusion Models