QRM: Quality-Aware Modulation for Stable Diffusion
Abstract
Modern diffusion models used for image generation, such as Stable Diffusion 3.5 (SD3.5), rely on timestep or prompt embeddings to modulate the strength of the denoising process in each timestep. While this per-timestep modulation communicates the current noise level, it does not provide any quality-aware information, which can lead to generated images lacking fidelity and prompt alignment. In this work, we propose the Quality Representation Module (QRM), a lightweight transformer that learns a quality-aware representation based on existing model inputs and can produce a quality vector $M_{qrm}$ for each diffusion timestep. This $M_{qrm}$ adjusts the adaptive LayerNorm modulation within SD3.5’s transformer blocks, injecting a quality-sensitive signal into the denoising parameters. The QRM introduces no major changes to the sampling schedule or diffusion backbone. Experiments include ablations on QRM training losses and architectures, as well as empirical results demonstrating consistent image quality improvements over SD3.5 across a range of evaluation metrics.
Read the paper
Related studies
- Learning from First-Generation Qualitative Approaches in the IS Discipline: An Evolutionary View and Some Implications for Authors and Evaluators (PART 1/2)
- Commentary on the clinical and preclinical dosage limits of interstitially administered magnetic fluids for therapeutic hyperthermia based on current practice and efficacy models
- Prosthesis: New Technological Opportunities and Innovative Biomedical Devices
- The history and future of neural modeling for cochlear implants
- Ryanodine receptors: waking up from refractoriness
- Novel Transgenic Rabbit Model Sheds Light on the Puzzling Role of Matrix Metalloproteinase-12 in Atherosclerosis
Go deeper in Evidence Hub
Evidence Hub adds a study data sheet (design, sample size, evidence level, key numbers), an AI summary and a chat grounded in the paper's text.