mlsanityData to 7 Oct 2026

Overcoming Common Flaws in the Evaluation of Selective Classification Systems

arXiv 2407.01032 ↗First posted 1 Jul 2024cs.LG cs.CV stat.MEcited by 31

Not on the map. Why?

Abstract (arXiv)

Selective Classification, wherein models can reject low-confidence predictions, promises reliable translation of machine-learning based classification systems to real-world scenarios such as clinical diagnostics. While current evaluation of these systems typically assumes fixed working points based on pre-defined rejection thresholds, methodological progress requires benchmarking the general performance of systems akin to the AUROC in standard classification. In this work, we define 5 requirements for multi-threshold metrics in selective classification regarding task alignment, interpretability, and flexibility, and show how current approaches fail to meet them. We propose the Area under the Generalized Risk Coverage curve (AUGRC), which meets all requirements and can be directly interpreted as the average risk of undetected failures. We empirically demonstrate the relevance of AUGRC on a comprehensive benchmark spanning 6 data sets and 13 confidence scoring functions. We find that the proposed metric substantially changes metric rankings on 5 out of the 6 data sets.

Key facts. Cited by 31 papers in the map (data through 7 Oct 2026). Builds on 39 of its 57 references that are in the map. First posted on arXiv 1 Jul 2024.

Builds on

39 of 57 references.

References · 39

  1. 2014Very Deep Convolutional Networks for Large-Scale Image Recognitioncited by 15,267
  2. 2017On Calibration of Modern Neural Networkscited by 4,860
  3. 2015Dropout as a Bayesian Approximation: Representing Model Uncertainty in Deep Learningcited by 4,416
  4. 2019Benchmarking Neural Network Robustness to Common Corruptions and Perturbationscited by 2,857
  5. 2020WILDS: A Benchmark of in-the-Wild Distribution Shiftscited by 1,366
  6. 2019Can You Trust Your Model's Uncertainty? Evaluating Predictive Uncertainty Under Dataset Shiftcited by 1,274
  7. 2017Selective Classification for Deep Neural Networkscited by 799
  8. 2023Scaling Vision Transformers to 22 Billion Parameterscited by 690
  9. 2019SelectiveNet: A Deep Neural Network with an Integrated Reject Optioncited by 333
  10. 2018Learning Confidence for Out-of-Distribution Detection in Neural Networkscited by 245
  11. 2020Pitfalls of In-Domain Uncertainty Estimation and Ensembling in Deep Learningcited by 208
  12. 2019Addressing Failure Prediction by Learning Model Confidencecited by 183
27 more
  1. 2020BREEDS: Benchmarks for Subpopulation Shiftcited by 155
  2. 2023Document Understanding Dataset and Evaluation (DUDE)cited by 132
  3. 2018Bias-Reduced Uncertainty Estimation for Deep Neural Classifierscited by 106
  4. 2022Plex: Towards Reliability using Pretrained Large Model Extensionscited by 102
  5. 2021Shifts: A Dataset of Real Distributional Shift Across Multiple Large-Scale Taskscited by 100
  6. 2019Deep Gamblers: Learning to Abstain with Portfolio Theorycited by 69
  7. 2021Optimal strategies for reject option classifierscited by 61
  8. 2019Methods and open-source toolkit for analyzing and visualizing challenge resultscited by 60
  9. 2022A Call to Reflect on Evaluation Practices for Failure Detection in Image Classificationcited by 51
  10. 2023Rethinking Confidence Calibration for Failure Predictioncited by 43
  11. 2023OpenMix: Exploring Outlier Samples for Misclassification Detectioncited by 38
  12. 2023What Can We Learn From The Selective Prediction And Uncertainty Estimation Performance Of 523 Imagenet Classifierscited by 35
  13. 2022Augmenting Softmax Information for Selective Classification with Out-of-Distribution Datacited by 27
  14. 2022Towards Better Selective Classificationcited by 21
  15. 2020Towards Improving Selective Prediction Ability of NLP Systemscited by 19
  16. 2021Disrupting Deep Uncertainty Estimation Without Harming Accuracycited by 19
  17. 2021A Unified Benchmark for the Unknown Detection Capability of Deep Neural Networkscited by 19
  18. 2021Measuring and Improving Model-Moderator Collaboration using Uncertainty Estimationcited by 18
  19. 2015Performance measures for classification systems with rejectioncited by 16
  20. 2023The Devil is in the Wrongly-classified Samples: Towards Unified Open-set Recognitioncited by 16
  21. 2023How to Fix a Broken Confidence Estimator: Evaluating Post-hoc Methods for Selective Classification with Deep Neural Networkscited by 15
  22. 2022AUC-based Selective Classificationcited by 13
  23. 2024Towards Understanding Why Label Smoothing Degrades Selective Classification and How to Fix Itcited by 11
  24. 2023Beyond Document Page Classification: Design, Datasets, and Challengescited by 10
  25. 2023Unified Classification and Rejection: A One-versus-All Frameworkcited by 9
  26. 2024Revisiting Confidence Estimation: Towards Reliable Failure Predictioncited by 9
  27. 2023Understanding Silent Failures in Medical Image Classificationcited by 5

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2024: 2; 2025: 6; 2026: 23.

31 papers.

Sort:

By area:

  1. 2024A Novel Characterization of the Population Area Under the Risk Coverage Curve (AURC) and Rates of Finite Sample Estimatorscited by 13
  2. 2026The Cost of Reasoning: Chain-of-Thought Induces Overconfidence in Vision-Language Modelscited by 5
  3. 2024Why context matters in VQA and Reasoning: Semantic interventions for VLM input modalitiescited by 4
  4. 2025What Does It Take to Build a Performant Selective Classifier?cited by 2
  5. 2025On the Reliability and Stability of Selective Methods in Malware Classification Taskscited by 1
  6. 2026Leveraging Data to Say No: Memory Augmented Plug-and-Play Selective Predictioncited by 1
  7. 2026Better than Average: Spatially-Aware Aggregation of Segmentation Uncertainty Improves Downstream Performancecited by 1
  8. 2026Confidence-Gated Robot Autonomy: When Does Uncertainty Actually Help?cited by 1
  9. 2026Does Your Wildfire Prediction Model Actually Work, or Just Score Well?cited by 1
  10. 2026Evaluating Epistemic Uncertainty: Beyond OOD Detection and Active Learningcited by 1
  11. 2026AnchorScore: A CLIP-Based Diagnostic of MLLM Annotation Difficultycited by 1
  12. 2025Trusted Uncertainty in Large Language Models: A Unified Framework for Confidence Calibration and Risk-Controlled Refusalcited by 0
19 more
  1. 2025Evaluating Medical LLMs by Levels of Autonomy: A Survey Moving from Benchmarks to Applicationscited by 0
  2. 2025Filtering instances and rejecting predictions to obtain reliable models in healthcarecited by 0
  3. 2025A Systematic Analysis of Out-of-Distribution Detection Under Representation and Training Paradigm Shiftscited by 0
  4. 2026Estimation of Confidence Bounds in Binary Classification using Wilson Score Kernel Density Estimationcited by 0
  5. 2026From Misclassifications to Outliers: Joint Reliability Assessment in Classificationcited by 0
  6. 2026The Confidence Gate Theorem: When Should Ranked Decision Systems Abstain?cited by 0
  7. 2026Less Data Approximates More: Earning Faithful Confidence in High-Stakes Domainscited by 0
  8. 2026CHASE: Competing Hypotheses for Ambiguity-Aware Selective Predictioncited by 0
  9. 2026A Regime Theory of Controller Class Selection for LLM Action Decisionscited by 0
  10. 2026ECUASn: A family of metrics for principled evaluation of uncertainty-augmented systemscited by 0
  11. 2026Proper Scoring Rules for Agentic Uncertainty Quantificationcited by 0
  12. 2026Valid Per-Field Selective Risk Control for Document Extraction: Three Failure Modes, a Validity Ladder, and When Conditioning Payscited by 0
  13. 2026One Score, Two Decisions: Selective Prediction on the Rare-Disease Tailcited by 0
  14. 2026ReliaGate: Reliability Routing for Low-Stakes Wearable Stress Predictioncited by 0
  15. 2026Verifiable abstention makes AI leak diagnosis accountable in urban water distribution networkscited by 0
  16. 2026Counterfactual Fragility Certificates: Exposing High-Confidence Brittleness under Structured Evidence Failurecited by 0
  17. 2026How Many Labels Does Model Choice Need? Certificates and Budgets for Selective Predictioncited by 0
  18. 2026SplatLabel: Pseudo-Labelling through 4D Gaussian Splattingcited by 0
  19. 2026CRC-Router: Risk-Constrained Routing for Medical Agentic AI Systemscited by 0