Research map / Label-efficient and robust learning / Recognition and long-tail
Active Testing: Sample-Efficient Model Evaluation
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 |
|---|---|---|---|
| Select test points using estimated loss and apply the bias-corrected estimator | Active testing · Synthetic datasets | squared error / cross-entropy estimation error | Helped |
| Use a retrained BNN surrogate instead of uniform sampling | Radial BNN · MNIST | median squared error | Helped |
| Use the original model's predictive entropy for acquisition (ablation) | Radial BNN · MNIST | median squared error | Helped |
| Use the original model's predictive entropy for acquisition (ablation) | ResNet-18 · Fashion-MNIST | median squared error | Hurt |
| Retrain the surrogate on acquired test labels (ablation) | Radial BNN · MNIST | median squared error | Helped |
| Retrain the surrogate on acquired test labels (ablation) | ResNet-18 · Fashion-MNIST | median squared error | Helped |
| Replace the Radial BNN surrogate with a random forest (ablation) | Radial BNN · MNIST | median squared error | No clear effect |
| Replace the ResNet surrogate with a random forest (ablation) | ResNet-18 · Fashion-MNIST | median squared error | Helped |
| Use a deep ensemble trained once instead of retraining a single ResNet surrogate (ablation) | ResNet-18 · Fashion-MNIST | median squared error | Helped |
| Increase ensemble diversity and surrogate capacity (ablation) | ResNet-18 · CIFAR-10 | squared error | Helped |
| Increase ensemble diversity rather than fidelity alone (ablation) | ResNet-18 · CIFAR-10 | squared error | Helped |
| Use mutual information instead of predictive entropy for acquisition (ablation) | Radial BNN · Fashion-MNIST | squared error | Hurt |
Closest papers by meaning
- Prediction-Powered Active Testing (NeurIPS 2026 · similarity 0.92)
- Scaling Up Active Testing to Large Language Models (NeurIPS 2025 · similarity 0.92)
- Active Statistical Inference (ICML 2024 · similarity 0.89)
- Contextual Active Model Selection (NeurIPS 2024 · similarity 0.87)
- Towards Cost-Effective Learning: A Synergy of Semi-Supervised and Active Learning (CVPR 2025 · similarity 0.87)
- Uncertainty Herding: One Active Learning Method for All Label Budgets (ICLR 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.