Research map / Label-efficient and robust learning / Recognition and long-tail
Classification Accuracy Score for Conditional 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 |
|---|---|---|---|
| Replace real ImageNet training examples with BigGAN-deep samples at 128×128 | ImageNet classifier training · ImageNet 128×128 | Top-5 accuracy: 88.79% -> 64.44% | Hurt |
| Replace real ImageNet training examples with HAM samples | ImageNet classifier training · ImageNet 128×128 | Top-5 accuracy: 64.44% -> 77.33% | Helped |
| Replace real ImageNet training examples with VQ-VAE-2 samples | ImageNet classifier training · ImageNet 256×256 | Top-5 accuracy: 65.92% -> 77.59% | Helped |
| Replace real ImageNet training examples with BigGAN-deep samples at 256×256 | ImageNet classifier training · ImageNet 256×256 | Top-5 accuracy: 91.47% -> 65.92% | Hurt |
| Use low-truncation BigGAN-deep samples to augment training data | ImageNet classifier training · ImageNet | Top-5 accuracy: 0.2% | Helped |
| Augment real ImageNet training data with other synthetic samples | ImageNet classifier training · ImageNet | classification accuracy | Hurt |
| Increase BigGAN-deep truncation across the tested range (ablation) | BigGAN-deep on ImageNet · ImageNet 256×256 | Top-5 accuracy: 13.24%; 28.68%; 32.88%; 45.01%; 56.68%; 62.97%; 65.92%; 64.37% | Mixed |
| Train on VQ-VAE-2 reconstructions rather than original images | ImageNet classifier training · ImageNet 256×256 | Top-5 accuracy: 91.47% -> 89.46% | Hurt |
| Use PixelCNN approximate classifier inference instead of exact Bayes inference | PixelCNN on CIFAR-10 · CIFAR-10 | accuracy: 60.05% -> 64.02% | Helped |
| Evaluate conditional generators using synthetic-trained downstream classification rather than only IS/FID | Generative model evaluation · ImageNet | Top-1 accuracy: correlation coefficient ... FID is 0.16, and IS -0.86 | Mixed |
Closest papers by meaning
- Fusing Conditional Submodular GAN and Programmatic Weak Supervision (AAAI 2024 · similarity 0.87)
- Generative Classifiers Avoid Shortcut Solutions (ICLR 2025 · similarity 0.87)
- Intriguing Properties of Generative Classifiers (ICLR 2024 · similarity 0.87)
- Direct Discriminative Optimization: Your Likelihood-Based Visual Generative Model is Secretly a GAN Discriminator (ICML 2025 · similarity 0.87)
- A Likelihood Based Approach to Distribution Regression Using Conditional Deep Generative Models (ICML 2025 · similarity 0.86)
- Taming the Tail in Class-Conditional GANs: Knowledge Sharing via Unconditional Training at Lower Resolutions (CVPR 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.