A Baseline for Detecting Misclassified and Out-of-Distribution Examples in Neural Networks

2016/10/31 by Dan Hendrycks, Kevin Gimpel · 15 citations
Computer Science · #cs.CV #cs.LG #cs.NE

paper · pdf

published as International Conference on Learning Representations 2017 · Published as a conference paper at ICLR 2017. 1 Figure in 1 Appendix. Minor changes from the previous version

arxiv created 2018/10/03 · arxiv updated 2018/10/04

Abstract

We consider the two related problems of detecting if an example is misclassified or out-of-distribution. We present a simple baseline that utilizes probabilities from softmax distributions. Correctly classified examples tend to have greater maximum softmax probabilities than erroneously classified and out-of-distribution examples, allowing for their detection. We assess performance by defining several tasks in computer vision, natural language processing, and automatic speech recognition, showing the effectiveness of this baseline across all. We then show the baseline can sometimes be surpassed, demonstrating the room for future research on these underexplored detection tasks.

Cited by

• Self-Aware Object Detection via Degradation Manifolds
• Intent Laundering: AI Safety Datasets Are Not What They Seem
• A Deployment Audit of Release-Side Risk in Conformal Triage under Prevalence Shift
• Filtered Reasoning Score: Evaluating Reasoning Quality on a Model's Most-Confident Traces
• Avoiding Leakage Poisoning: Concept Interventions Under Distribution Shifts
• Enhancing System Self-Awareness and Trust of AI: A Case Study in Trajectory Prediction and Planning
• GUARD: Grounding Uncertainty and Ablation-Based Risk Detection for Diffusion-Based VLAs
• Can We Ignore Labels In Out of Distribution Detection?
• TrustLoRA: Low-Rank Adaptation for Failure Detection under Out-of-distribution Data
• PVLM: Parsing-Aware Vision Language Model with Dynamic Contrastive Learning for Zero-Shot Deepfake Attribution
• Monitor and Recover: A Paradigm for Future Research on Distribution Shift in Learning-Enabled Cyber-Physical Systems
• Safety Monitoring for Learning-Enabled Cyber-Physical Systems in Out-of-Distribution Scenarios
• Bounded and Uniform Energy-based Out-of-distribution Detection for Graphs
• On the calibration of Just-in-time Defect Prediction
• Efficient Contrastive Decoding with Probabilistic Hallucination Detection - Mitigating Hallucinations in Large Vision Language Models -

Related

• Detecting Out-of-Distribution Examples with In-distribution Examples and Gram Matrices
• Why Should we Combine Training and Post-Training Methods for Out-of-Distribution Detection?
• Detecting Misclassification Errors in Neural Networks with a Gaussian Process Model
• Understanding the Feature Norm for Out-of-Distribution Detection
• Utilizing Network Properties to Detect Erroneous Inputs