Research map / Learning theory and algorithms / Prediction and decision losses
A Model-Agnostic Heuristics for Selective Classification
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 |
|---|---|---|---|
| Use cross-fitting and a full-sample/subsample quantile threshold rather than a validation-set empirical quantile | LightGBM selective classification · Tabular datasets | selective accuracy: 27/48 cases | Helped |
| Use cross-fitting and a full-sample/subsample quantile threshold rather than a validation-set empirical quantile | LightGBM selective classification · Tabular datasets | empirical coverage closeness to target: 19/48 cases vs PLUGIN 11/48 | Helped |
| Use SCROSS instead of end-to-end SelectiveNet or SAT | Selective classification · Tabular datasets | selective accuracy | Helped |
| Use SCROSS instead of end-to-end SelectiveNet or SAT | Selective classification · CatsVsDogs | selective accuracy | Hurt |
| Increase the number of cross-fitting folds (ablation) | SCROSS · CatsVsDogs | empirical coverage | Hurt |
| Increase the number of cross-fitting folds (ablation) | SCROSS · Tabular datasets | selective accuracy | No clear effect |
| Use K=7 or K=10 folds instead of K=5 (ablation) | SCROSS · CSDS1 and CSDS2 | empirical coverage | Helped |
| Replace LightGBM with a simpler Logistic base classifier (ablation) | SCROSS · Tabular datasets | empirical coverage: 25/42 cases | Helped |
| Replace the base classifier with a more complex model (ablation) | SCROSS · Tabular datasets | selective accuracy: XGBoost 30/42 cases | Helped |
| Use cross-fitting to avoid setting aside a validation set | SCROSS · Tabular and image datasets | elapsed training time: K times the elapsed time of PLUG-IN | Hurt |
Closest papers by meaning
- Overcoming Common Flaws in the Evaluation of Selective Classification Systems (NeurIPS 2024 · similarity 0.85)
- Distribution-Specific Agnostic Conditional Classification With Halfspaces (ICLR 2025 · similarity 0.84)
- What Does It Take to Build a Performant Selective Classifier? (NeurIPS 2025 · similarity 0.84)
- Revisiting Agnostic Boosting (NeurIPS 2025 · similarity 0.83)
- Learning with Selectively Labeled Data from Multiple Decision-makers (ICML 2025 · similarity 0.83)
- Sample Complexity of Algorithm Selection Using Neural Networks and Its Applications to Branch-and-Cut (NeurIPS 2024 · 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.