EMAG: Self-Rectifying Diffusion Sampling with Exponential Moving Average Guidance
Published:Dec 19, 2025 07:36
•1 min read
•ArXiv
Analysis
The article introduces a new method called EMAG for diffusion sampling. The core idea involves self-rectification and the use of exponential moving average guidance. This suggests an improvement in the efficiency or quality of diffusion models, potentially addressing issues related to sampling instability or slow convergence. The source being ArXiv indicates this is a research paper, likely detailing the technical aspects, experimental results, and comparisons to existing methods.
Key Takeaways
Reference
“”