Research map / Image generation / GANs
Rejection Sampling IMLE: Designing Priors for Better Few-Shot Image Synthesis
Chirag Vashist, Shichong Peng, Ke Li
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 |
|---|---|---|---|
| train with rejection-sampled prior (reject samples within epsilon of any data point) instead of Gaussian prior | IMLE (AdaIMLE) · Obama | FID: 25.0 -> 14.0 | Improved |
| train with rejection-sampled prior | AdaIMLE · FFHQ-100 | FID: 33.2 -> 12.9 | Improved |
| train with rejection-sampled prior | AdaIMLE · Skulls | recall: 0.32 -> 0.65 | Improved |
| train with rejection-sampled prior | AdaIMLE · Shells | recall: 0.62 -> 0.59 | No clear effect |
| increase rejection radius epsilon beyond about 0.15 (ablation) | RS-IMLE · FFHQ-100 | FID: 12.91 -> 14.8 | Hurt |
| decrease epsilon from 0.15 to 0.12 (ablation) | RS-IMLE · Animal-Face Cat | FID: 15.96 -> 17.29 | Hurt |
| reduce latent dimension from 1024 to 256 (fewer parameters) (ablation) | RS-IMLE · Shells | FID: 55.4 -> 71.7 | Hurt |
| reduce latent dimension to 512 (19M params) (ablation) | RS-IMLE · Shells | FID: 55.4 -> 52.9 | No clear effect |
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.
- FakeCLR — FakeCLR: Exploring Contrastive Learning for Solving Latent Discontinuity in Data-Efficient GANs (ECCV 2022) · better on 1 of 1 shared result
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
- Adaptive IMLE for Few-shot Pretraining-free Generative Modelling (ICML 2023 · similarity 0.95)
- You Only Sample Once: Taming One-Step Text-to-Image Synthesis by Self-Cooperative Diffusion GANs (ICLR 2025 · similarity 0.84)
- Learning Energy-Based Models by Cooperative Diffusion Recovery Likelihood (ICLR 2024 · similarity 0.83)
- Inductive Moment Matching (ICML 2025 · similarity 0.83)
- Revisiting Diffusion Models: From Generative Pre-training to One-Step Generation (ICML 2025 · similarity 0.83)
- Improved Distribution Matching Distillation for Fast Image Synthesis (NeurIPS 2024 · similarity 0.83)
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.
