Research map / Learning theory and algorithms / Prediction and decision losses
Front-Loaded Robust Conformal Prediction: Heavy Calibration, Minimal Test-Time Cost
Soroush H. Zargarbashi, Mohammad Sadegh Akhondzadeh, Aleksandar Bojchevski
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 |
|---|---|---|---|
| replace RCP1 with FLoR(1) (heavy calibration sampling) | RCP1 · ImageNet ResNet r=0.50 | avg set size: 144.226 -> 24.076 | Improved |
| replace RCP1 with FLoR(1) | RCP1 · ImageNet ResNet r=0.25 | avg set size: 45.019 -> 15.905 | Improved |
| replace RCP1 with FLoR(1) | RCP1 · ImageNet Diffusion+ViT r=0.50 | avg set size: 197.633 -> 20.765 | Improved |
| replace RCP1 with FLoR(1) | RCP1 · CIFAR-10 ResNet r=0.50 | avg set size: 5.413 -> 4.165 | Improved |
| replace RCP1 with FLoR(1) | RCP1 · CIFAR-10 Diffusion+ViT r=0.50 | avg set size: 8.935 -> 6.453 | Improved |
| replace RCP1 with FLoR(1) at zero radius | RCP1 · ImageNet ResNet r=0.00 | avg set size: 8.605 -> 12.451 | Hurt |
| replace RCP1 with FLoR(1) at zero radius | RCP1 · CIFAR-10 ResNet r=0.00 | avg set size: 2.470 -> 2.680 | Hurt |
| disable finite-sample correction in FLoR(1) (APS score) (ablation) | FLoR(1) · ImageNet/CIFAR APS, ResNet | avg set size | Improved |
| increase calibration-time MC sample rate (ablation) | FLoR(1) · CIFAR-10 ResNet r=0.5 | avg set size | Improved |
| calibrate FLoR(k) for majority vote at low test-time sample rate (21 to 101) | BinCP · CIFAR-10 ResNet | avg set size | Improved |
| use plain majority vote of test-time sets without calibration awareness | RCP1/FLoR(1) · coverage guarantee | coverage: 1 - alpha -> 1 - 2 alpha | Hurt |
| increase smoothing sigma (ablation) | FLoR(1) · ImageNet/CIFAR-10 | avg set size | Hurt |
Closest papers by meaning
- One Sample is Enough to Make Conformal Prediction Robust (NeurIPS 2025 · similarity 0.93)
- Robust Conformal Prediction with a Single Binary Certificate (ICLR 2025 · similarity 0.89)
- Provably Robust Conformal Prediction with Improved Efficiency (ICLR 2024 · similarity 0.89)
- Robust Conditional Conformal Prediction via Branched Normalizing Flow (NeurIPS 2026 · similarity 0.88)
- Co-optimization for Adaptive Conformal Prediction (NeurIPS 2026 · similarity 0.88)
- Exploring the Noise Robustness of Online Conformal Prediction (NeurIPS 2025 · similarity 0.88)
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.
