Research map / Image generation / Autoregressive and latent generation
Improved Precision and Recall Metric for Assessing 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 |
|---|---|---|---|
| Use the improved k-NN precision and recall metric instead of relying on FID alone | Generative image evaluation · StyleGAN and BigGAN | precision and recall interpretability | Helped |
| Remove the minibatch standard deviation layer (ablation) | StyleGAN · FFHQ | precision | Helped |
| Remove the minibatch standard deviation layer (ablation) | StyleGAN · FFHQ | recall | Hurt |
| Reduce the R1 regularization gamma parameter by 100× (ablation) | StyleGAN · FFHQ | precision | Helped |
| Reduce the R1 regularization gamma parameter by 100× (ablation) | StyleGAN · FFHQ | recall | Hurt |
| Disable progressive growing (ablation) | StyleGAN · FFHQ | precision | Hurt |
| Disable progressive growing (ablation) | StyleGAN · FFHQ | recall | Hurt |
| Randomly translate discriminator inputs by −16…16 pixels (ablation) | StyleGAN · FFHQ | precision | Helped |
| Randomly translate discriminator inputs by −16…16 pixels (ablation) | StyleGAN · FFHQ | recall | Hurt |
| Randomly translate discriminator inputs by −16…16 pixels (ablation) | StyleGAN · FFHQ | FID | Helped |
| Disable instance normalization in AdaIN (ablation) | StyleGAN · FFHQ | recall | Helped |
| Reject latent vectors by density rather than by distance (ablation) | StyleGAN truncation · FFHQ | precision-recall tradeoff | Helped |
Closest papers by meaning
- Efficient Precision and Recall Metrics for Assessing Generative Models using Hubness-aware Sampling (ICML 2024 · similarity 0.91)
- A Unifying Information-theoretic Perspective on Evaluating Generative Models (AAAI 2025 · similarity 0.89)
- GenAI Arena: An Open Evaluation Platform for Generative Models (NeurIPS 2024 · similarity 0.87)
- Anomaly Score: Evaluating Generative Models and Individual Generated Images based on Complexity and Vulnerability (CVPR 2024 · similarity 0.87)
- Statistically Optimal Generative Modeling with Maximum Deviation from the Empirical Distribution (ICML 2024 · similarity 0.87)
- FoMo: Forking Moment in Generative Trajectory as a Perceptual Distance (NeurIPS 2026 · 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.