Research map / Label-efficient and robust learning / Recognition and long-tail
How Faithful is your Synthetic Data? Sample-level Metrics for Evaluating and Auditing 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 |
|---|---|---|---|
| Audit ADS-GAN samples by rejecting those below precision or authenticity thresholds (ablation) | ADS-GAN on SIVEP-GRIPE · SIVEP-GRIPE | predictive model AUC-ROC: from 0.76 to 0.78 | Helped |
| Audit ADS-GAN samples by rejecting those below precision or authenticity thresholds (ablation) | ADS-GAN on SIVEP-GRIPE · SIVEP-GRIPE | Authenticity: 0.5905 -> 1.0000 | Helped |
| Audit ADS-GAN samples by rejecting those below precision or authenticity thresholds (ablation) | ADS-GAN on SIVEP-GRIPE · SIVEP-GRIPE | IPα: 0.7744 -> 0.9994 | Helped |
| Audit hamada samples by rejecting those below precision or authenticity thresholds (ablation) | hamada on AmsterdamUMCdb · AmsterdamUMCdb | Authenticity: 0.3572 -> 1.0 | Helped |
| Audit hamada samples by rejecting those below precision or authenticity thresholds (ablation) | hamada on AmsterdamUMCdb · AmsterdamUMCdb | IPα: 0.5482 -> 0.9976 | Helped |
| Audit hamada samples by rejecting those below precision or authenticity thresholds (ablation) | hamada on AmsterdamUMCdb · AmsterdamUMCdb | IRβ: 0.0850 -> 0.0334 | Hurt |
| Increase ADS-GAN privacy regularization λ | ADS-GAN · SIVEP-GRIPE | Authenticity | Helped |
| Increase ADS-GAN privacy regularization λ | ADS-GAN · SIVEP-GRIPE | precision | Hurt |
| Use DDPM instead of StyleGAN2-ADA | CIFAR-10 image generation · CIFAR-10 | IPα: 0.858 -> 0.921 | Helped |
| Use DDPM instead of StyleGAN2-ADA | CIFAR-10 image generation · CIFAR-10 | IRβ: 0.663 -> 0.501 | Hurt |
| Increase mode dropping by replacing deleted digits 1–9 with digit 0 | CGAN-generated MNIST · MNIST | IRβ | Hurt |
| Increase mode dropping by replacing deleted digits 1–9 with digit 0 | CGAN-generated MNIST · MNIST | IPα | No clear effect |
Closest papers by meaning
- Position: All Current Generative Fidelity and Diversity Metrics are Flawed (ICML 2025 · similarity 0.93)
- A Unifying Information-theoretic Perspective on Evaluating Generative Models (AAAI 2025 · similarity 0.90)
- Debiasing Synthetic Data Generated by Deep Generative Models (NeurIPS 2024 · similarity 0.87)
- Efficient Precision and Recall Metrics for Assessing Generative Models using Hubness-aware Sampling (ICML 2024 · similarity 0.87)
- On the Stability of Iterative Retraining of Generative Models on their own Data (ICLR 2024 · similarity 0.87)
- Image Generation Diversity Issues and How to Tame Them (CVPR 2025 · 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.