Research map / Learning theory and algorithms / Prediction and decision losses
Quantifying and Improving Adaptivity in Conformal Prediction Through Input Transformations
Sooyong Jang, Insup Lee
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 |
|---|---|---|---|
| apply difficulty-binned group-conditional CP (O-LAC) instead of LAC | LAC · ImageNet val alpha=0.10 (avg of 9 classifiers) | T-SS: 0.543 -> 0.737 | Improved |
| apply difficulty-binned group-conditional CP (O-LAC) instead of LAC | LAC · ImageNet val alpha=0.10 (avg of 9 classifiers) | T-CV: 0.304 -> 0.050 | Improved |
| apply difficulty-binned group-conditional CP (O-SAPS) instead of SAPS | SAPS · ImageNet val alpha=0.10 | T-CV: 0.148 -> 0.045 | Improved |
| apply difficulty-binned group-conditional CP (O-SAPS) instead of SAPS | SAPS · ImageNet val alpha=0.10 | T-SS: 0.673 -> 0.720 | Improved |
| apply O-LAC vs APS on T-CV | APS · ImageNet val alpha=0.10 | T-CV: 0.065 -> 0.050 | No clear effect |
| apply O-LAC vs APS on T-SS for ResNet50 | APS · ImageNet val ResNet50 alpha=0.10 | T-SS: 0.210 -> 0.835 | Improved |
| use O-LAC at alpha=0.05 instead of LAC | LAC · ImageNet val alpha=0.05 | T-SS: 0.678 -> 0.743 | Improved |
| use more accurate base classifier (ViT-H-14 vs ResNet18) | O-LAC · ImageNet val alpha=0.10 | T-SS: 0.926 -> 0.627 | Hurt |
| increase target miscoverage alpha to 0.15-0.20 | T-SS metric · ImageNet val | T-SS | Hurt |
| bin by difficulty and run per-bin thresholds (group-conditional CP) | O-LAC · ImageNet val | average set size | Mixed |
| use O-CPA/O-PAC instead of baselines for interval width vs error | PAC · Visual acuity alpha=0.30, EfficientNet-V2-S | R2: 0.89 -> 0.92 | Improved |
| use O-CPA instead of CP-A for interval width vs error | CP-A · Visual acuity alpha=0.30, EfficientNet-V2-S | R2: 0.87 -> 0.87 | No clear effect |
Closest papers by meaning
- Hierarchical Conformal Classification (NeurIPS 2026 · similarity 0.87)
- Co-optimization for Adaptive Conformal Prediction (NeurIPS 2026 · similarity 0.87)
- Multi-model Ensemble Conformal Prediction in Dynamic Environments (NeurIPS 2024 · similarity 0.87)
- Valid Selection among Conformal Sets (NeurIPS 2025 · similarity 0.87)
- Scaling Full Conformal Image Classifiers (NeurIPS 2026 · similarity 0.86)
- Conformal Prediction for Zero-Shot Models (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.
