Research map / Label-efficient and robust learning / Recognition and long-tail
Black-Box Ripper: Copying black-box models using generative evolutionary algorithms
Antonio Barbalau, Adrian Cosma, Radu Tudor Ionescu, Marius Popescu
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 |
|---|---|---|---|
| add evolutionary search to GAN-generated samples (ablation) | Black-Box Ripper (SNGAN) · Fashion-MNIST VGG-16 | accuracy (%): 87.6 -> 90.0 | Improved |
| add evolutionary search to GAN-generated samples (ablation) | Black-Box Ripper (SNGAN) · Fashion-MNIST LeNet->Half LeNet | accuracy (%): 80.0 -> 82.2 | Improved |
| add evolutionary search to VAE-generated samples (ablation) | Black-Box Ripper (VAE) · Fashion-MNIST VGG-16 | accuracy (%): 78.3 -> 86.1 | Improved |
| add evolutionary search to VAE-generated samples (ablation) | Black-Box Ripper (VAE) · Fashion-MNIST LeNet->Half LeNet | accuracy (%): 73.1 -> 78.8 | Improved |
| add evolutionary search to ProGAN samples (ablation) | Black-Box Ripper · 10 Monkey Species (CelebA-HQ proxy) | accuracy (%): 57.3 -> 69.4 | Improved |
| add evolutionary search to cGAN samples (ablation) | Black-Box Ripper · 10 Monkey Species (ImageNet Cats and Dogs proxy) | accuracy (%): 65.0 -> 71.6 | Improved |
| use GAN-generated samples with evolutionary search instead of raw proxy data (Knockoff Nets) | Black-Box Ripper · 10 Monkey Species (CelebA-HQ proxy) | accuracy (%): 54.7 -> 69.4 | Improved |
| reduce proxy dataset classes from 90 to 6 (gain over DeGAN grows) | Black-Box Ripper vs DeGAN · CIFAR-10 | accuracy (%): 79.0 vs 80.5 (90 classes); 69.9 vs 59.5 (6 classes) | Mixed |
Closest papers by meaning
- Posterior Inference in Latent Space for Scalable Constrained Black-box Optimization (NeurIPS 2026 · similarity 0.85)
- Where's the Liability in the Generative Era? Recovery-based Black-Box Detection of AI-Generated Content (CVPR 2025 · similarity 0.85)
- Knockoff Nets: Stealing Functionality of Black-Box Models (CVPR 2019 · similarity 0.84)
- Few-Shot Learning via Repurposing Ensemble of Black-Box Models (AAAI 2024 · similarity 0.84)
- EvolveDirector: Approaching Advanced Text-to-Image Generation with Large Vision-Language Models (NeurIPS 2024 · similarity 0.84)
- Posterior Inference with Diffusion Models for High-dimensional Black-box Optimization (ICML 2025 · 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.
