Research map / Image generation / Fast diffusion sampling
MMD Guidance: Training-Free Distribution Adaptation for Diffusion Models via Maximum Mean Discrepancy Guidance
Matina Mahdizadeh Sani, Nima Jamali, Mohammad Jalali, Farzan Farnia
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 MMD guidance to unguided LDM | LDM · FFHQ sunglasses user | FD: 1220.5 -> 693.0 | Improved |
| add MMD guidance to unguided LDM | LDM · FFHQ reading-glasses user | FD: 702.9 -> 574.3 | Improved |
| use MMD guidance instead of Domain Guidance | LDM · FFHQ sunglasses user | FD: 710.6 -> 693.0 | Improved |
| use classifier guidance instead of no guidance | LDM · FFHQ sunglasses user | FD: 1220.5 -> 1195.9 | No clear effect |
| add MMD guidance with product kernel | SDXL · Prompt-aware style dataset | FD: 1953.8 -> 1674.4 | Improved |
| add MMD guidance with product kernel | PixArt · Prompt-aware style dataset | FD: 1358.4 -> 1060.6 | Improved |
| add MMD guidance (runtime overhead) | SDXL · Runtime, 500 samples | cumulative time: 3142 -> 3447 | Hurt |
| increase reference samples from 10 to 200 (ablation) | MMD Guidance · Prompt-aware style dataset | FD: 1805.92 -> 1674.45 | Improved |
| increase reference samples from 10 to 100 (ablation) | MMD Guidance · GMM large gap (r=20) | FD: 123.22 -> 108.61 | Improved |
| increase reference samples from 10 to 100 (ablation) | MMD Guidance · GMM shifted target | FD: 0.60 -> 0.03 | Improved |
| keep the intra-batch diversity term (ablation) | MMD Guidance · synthetic GMM and FFHQ | mode coverage | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- Domain Guidance — Domain Guidance: A Simple Transfer Approach for a Pre-trained Diffusion Model (ICLR 2025) · better on 4 of 4 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
- Domain Guidance: A Simple Transfer Approach for a Pre-trained Diffusion Model (ICLR 2025 · similarity 0.91)
- Token Perturbation Guidance for Diffusion Models (NeurIPS 2025 · similarity 0.89)
- Transfer Learning for Diffusion Models (NeurIPS 2024 · similarity 0.89)
- TFG: Unified Training-Free Guidance for Diffusion Models (NeurIPS 2024 · similarity 0.89)
- No Training, No Problem: Rethinking Classifier-Free Guidance for Diffusion Models (ICLR 2025 · similarity 0.89)
- Applying Guidance in a Limited Interval Improves Sample and Distribution Quality in Diffusion Models (NeurIPS 2024 · similarity 0.89)
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.
