Research map / Image generation / Fast diffusion sampling
Initialization is Half the Battle: Generating Diverse Images from a Guidance Potential Posterior
Xiang Li, Dianbo Liu, Kenji Kawaguchi
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 Langevin noise injection (ablation) | DivIn · ImageNet SD v1.4 | recall: 0.541 -> 0.569 | Improved |
| add Langevin noise injection (ablation) | DivIn · ImageNet SD v1.4 | Vendi score: 4.534 -> 4.688 | Improved |
| add the Gaussian prior term (ablation) | DivIn · ImageNet SD v1.4 | recall: 0.557 -> 0.569 | Improved |
| add the Gaussian prior term (ablation) | DivIn · ImageNet SD v1.4 | FID: 16.121 -> 16.158 | Hurt |
| add DivIn initialization to base model | SD v1.4 · ImageNet | Vendi score: 4.265 -> 4.688 | Improved |
| add DivIn initialization to base model | SD v1.4 · ImageNet | FID: 16.696 -> 16.158 | Improved |
| add DivIn to Particle Guidance | PG, SD v1.4 · ImageNet | FID: 17.520 -> 17.942 | Hurt |
| add DivIn to CADS | CADS, SD v1.4 · ImageNet | recall: 0.528 -> 0.553 | Improved |
| add DivIn to Interval Guidance | IG, SD v1.4 · ImageNet | Vendi score: 4.585 -> 4.729 | Improved |
| add DivIn to IG on SD v3.5 | IG, SD v3.5 · general prompts | ImageReward: 0.521 -> 0.501 | Hurt |
| use DivIn instead of SAIL | SD v1.4 · ImageNet | Vendi score: 4.549 -> 4.688 | Improved |
| increase Langevin steps K to 3 without noise or prior (ablation) | DivIn variants · ImageNet | FID: 17.53 and 17.36 vs 15.98 for full DivIn | Hurt |
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.
- CADS — CADS: Unleashing the Diversity of Diffusion Models through Condition-Annealed Sampling (ICLR 2024) · 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
- Particle Guidance: non-I.I.D. Diverse Sampling with Diffusion Models (ICLR 2024 · similarity 0.88)
- Consistency Posterior Sampling for Diverse Image Synthesis (CVPR 2025 · similarity 0.87)
- The Crystal Ball Hypothesis in diffusion models: Anticipating object positions from initial noise (ICLR 2025 · similarity 0.87)
- Greed is Good: A Unifying Perspective on Guided Generation (NeurIPS 2025 · similarity 0.86)
- SPARKE: Scalable Prompt-Aware Diversity and Novelty Guidance in Diffusion Models via RKE Score (NeurIPS 2025 · similarity 0.86)
- Theoretical insights for diffusion guidance: A case study for Gaussian mixture models (ICML 2024 · similarity 0.86)
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.
