Research map / Image generation / Autoregressive and latent generation
Diversity vs. Recognizability: Human-like generalization in one-shot generative models
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| increase number of context samples from 2 to 20 (ablation) | VAE-NS · Omniglot | sample diversity: 2.4 -> 1.2 | Hurt |
| increase number of context samples from 2 to 20 (ablation) | VAE-NS · Omniglot | classification accuracy: 53 -> 96 | Helped |
| vary number of attentional steps from 20 to 90 (ablation) | VAE-STN · Omniglot | recognizability | Mixed |
| vary number of attentional steps from 20 to 90 (ablation) | VAE-STN · Omniglot | sample diversity | Mixed |
| increase beta in ELBO (ablation) | VAE-NS and VAE-STN · Omniglot | sample diversity | Hurt |
| increase beta in ELBO (ablation) | VAE-STN · Omniglot | recognizability: 2.5 | Mixed |
| increase beta in ELBO (ablation) | VAE-NS · Omniglot | recognizability: 3 | Mixed |
| change latent space size (ablation) | DAGAN-UN, DAGAN-RN, VAE-NS, VAE-STN · Omniglot | diversity and recognizability | No clear effect |
| use U-Net skip connections (DAGAN-UN vs DAGAN-RN) (ablation) | DAGAN · Omniglot | recognizability and diversity | Mixed |
| use supervised instead of unsupervised features for diversity (ablation) | diversity measure · Omniglot | diversity correlation: 0.86 | No clear effect |
Closest papers by meaning
- Intriguing Properties of Generative Classifiers (ICLR 2024 · similarity 0.88)
- Statistically Optimal Generative Modeling with Maximum Deviation from the Empirical Distribution (ICML 2024 · similarity 0.87)
- Image Generation Diversity Issues and How to Tame Them (CVPR 2025 · similarity 0.87)
- GenAI Arena: An Open Evaluation Platform for Generative Models (NeurIPS 2024 · similarity 0.86)
- Minimalist Concept Erasure in Generative Models (ICML 2025 · similarity 0.86)
- Latent Representation Matters: Human-like Sketches in One-shot Drawing Tasks (NeurIPS 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.