mlsanityData to 14 Oct 2026

QSilk: Micrograin Stabilization and Adaptive Quantile Clipping for Detail-Friendly Latent Diffusion

arXiv 2510.15761 ↗First posted 17 Oct 2025cs.CV cs.LGcited by 0

On the map: Diffusion models → Diffusion foundations and sampling

Abstract (arXiv)

We present QSilk, a lightweight, always-on stabilization layer for latent diffusion that improves high-frequency fidelity while suppressing rare activation spikes. QSilk combines (i) a per-sample micro clamp that gently limits extreme values without washing out texture, and (ii) Adaptive Quantile Clip (AQClip), which adapts the allowed value corridor per region. AQClip can operate in a proxy mode using local structure statistics or in an attention entropy guided mode (model confidence). Integrated into the CADE 2.5 rendering pipeline, QSilk yields cleaner, sharper results at low step counts and ultra-high resolutions with negligible overhead. It requires no training or fine-tuning and exposes minimal user controls. We report consistent qualitative improvements across SD/SDXL backbones and show synergy with CFG/Rescale, enabling slightly higher guidance without artifacts.

Key facts. No citing papers in the map yet (data through 14 Oct 2026). Placed in Diffusion foundations and sampling, one of 14 directions in Diffusion models. Builds on 8 of its 12 references that are in the map. First posted on arXiv 17 Oct 2025.

Builds on

8 of 12 references.

Closest ancestors · 5

Same direction, newest first

  1. 2025Classifier-free Guidance with Adaptive Scaling14 Feb
  2. 2024Rethinking the Spatial Inconsistency in Classifier-Free Diffusion Guidance8 Apr
  3. 2022Classifier-Free Diffusion Guidance26 Jul
  4. 2021High-Resolution Image Synthesis with Latent Diffusion Models20 Dec
  5. 2020Denoising Diffusion Probabilistic Models19 Jun

Other references · 3

  1. 2022Photorealistic Text-to-Image Diffusion Models with Deep Language Understandingcited by 6,355
  2. 2023Active Visual Exploration Based on Attention-Map Entropycited by 13
  3. 2025Understanding Attention Mechanism in Video Diffusion Modelscited by 7

Built on by

Not cited yet.

Similar, not cited

Close in wording, no citation either way

  1. 2026From Sketch to Fresco: Efficient Diffusion Transformer with Progressive Resolution
  2. 2024FAM Diffusion: Frequency and Attention Modulation for High-Resolution Image Generation with Stable Diffusion
  3. 2025Ctrl-Z Sampling: Scaling Diffusion Sampling with Controlled Random Zigzag Explorations
  4. 2025CADE 2.5 - ZeResFDG: Frequency-Decoupled, Rescaled and Zero-Projected Guidance for SD/SDXL Latent Diffusion Models
  5. 2026Q-Drift: Quantization-Aware Drift Correction for Diffusion Model Sampling
  6. 2025DiffPro: Joint Timestep and Layer-Wise Precision Optimization for Efficient Diffusion Inference
  7. 2024CutDiffusion: A Simple, Fast, Cheap, and Strong Diffusion Extrapolation Method
  8. 2026CentriQ: Calibration-Free Quantization of Diffusion Transformers via Exact Mean Centering