Research map / Image generation / Fast diffusion sampling
EMAG: Self-Rectifying Diffusion Sampling with Exponential Moving Average Guidance
ANKIT YADAV, Huy Ta, Lingqiao Liu
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add EMAG to CFG | CFG · SD3 COCO 2014 val | HPS: 29.22 -> 29.76 | Improved |
| add EMAG to APG | APG · SD3 COCO 2014 val | HPS: 29.45 -> 29.79 | Improved |
| add EMAG to CADS | CADS · SD3 COCO 2014 val | HPS: 28.36 -> 28.86 | Mixed |
| add EMAG to ERG | ERG · SD3 COCO 2014 val | HPS: 29.35 -> 29.56 | Improved |
| add EMAG to S2 | S2 · SD3 COCO 2014 val | HPS: 29.15 -> 29.51 | Improved |
| add EMAG to SEG + CFG | SEG + CFG · SD3 COCO 2014 val | HPS: 29.25 -> 29.29 | No clear effect |
| restrict EMAG to one adaptive layer instead of all layers (ablation) | EMAG · SD3 layer ablation | FID: 35.60 -> 28.52 | Improved |
| restrict EMAG to adaptive layer instead of all layers (ablation) | EMAG · SD3 layer ablation | HPS: 28.80 -> 29.60 | Improved |
| use EMAG-Q (quantized variant) instead of EMAG | EMAG · SD3 COCO 2014 val | inference overhead: +213% -> +95% | Improved |
| use EMAG-Q instead of EMAG for quality | EMAG · SD3 COCO 2014 val | FID and HPS: +0.63 FID, -0.12 HPS | Hurt |
| use EMAG guidance instead of no guidance, unconditional | SD3 · SD3 unconditional | FID: 101.94 -> 74.98 | Improved |
| use EMAG guidance instead of no guidance, DiT-XL/2 256 | DiT-XL/2 · ImageNet-256 unconditional | FID: 52.76 -> 29.85 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 0 of 1 on most shared results.
- CADS — CADS: Unleashing the Diversity of Diffusion Models through Condition-Annealed Sampling (ICLR 2024) · better on 1 of 2 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Normalized Attention Guidance: Universal Negative Guidance for Diffusion Models (NeurIPS 2025 · similarity 0.91)
- Adaptive Guidance: Training-free Acceleration of Conditional Diffusion Models (AAAI 2025 · similarity 0.90)
- Eliminating Oversaturation and Artifacts of High Guidance Scales in Diffusion Models (ICLR 2025 · similarity 0.90)
- Applying Guidance in a Limited Interval Improves Sample and Distribution Quality in Diffusion Models (NeurIPS 2024 · similarity 0.90)
- Feedback Guidance of Diffusion Models (NeurIPS 2025 · similarity 0.90)
- Tortoise and Hare Guidance: Accelerating Diffusion Model Inference with Multirate Integration (NeurIPS 2025 · similarity 0.90)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
