Medix: Out-of-Distribution Detection from Unlabeled Wild Data via Robust Gradient Statistics

2025/10/07 by Mohammed Fadhil Abbas, Abbas, Momin, Falahati, Ali +4 · 2 citations
Computer Science · #Anomaly Detection Techniques and Applications #Artificial Intelligence (cs.AI) #FOS: Computer and information sciences #FOS: Mathematics #Machine Learning (cs.LG) #Machine Learning (stat.ML) #Optimization and Control (math.OC)

paper · pdf · doi:10.48550/arxiv.2510.06505

openalex publication_date 2025/10/07 · openalex created_date 2025/10/18 · openalex updated_date 2026/07/28

Abstract

Out-of-distribution (OOD) detection plays a crucial role in ensuring the robustness of machine learning systems deployed in real-world applications. Recent approaches have explored the use of unlabeled data, showing potential for enhancing OOD detection capabilities. However, effectively utilizing unlabeled in-the-wild data remains challenging due to the mixed nature of both in-distribution (InD) and OOD samples. The lack of a distinct set of OOD samples complicates the task of training an optimal OOD classifier. In this work, we introduce Medix, a novel framework designed to identify potential outliers from unlabeled data using the median-based robust gradient statistics. We use the median because it provides a stable estimate of the central tendency, as an OOD detection mechanism, due to its robustness against noise and outliers. Using these identified outliers, along with labeled InD data, we train a robust OOD classifier. From a theoretical perspective, we derive error bounds that demonstrate Medix achieves a low error rate. Empirical results further substantiate our claims, as Medix outperforms existing methods across the board in open-world settings.

Citations

• GOOD: Training-Free Guided Diffusion Sampling for Out-of-Distribution Detection
• LoD: Loss-difference OOD Detection by Intentionally Label-Noisifying Unlabeled Wild Data
• Out-of-Distribution Detection using Synthetic Data Generation
• SCOMatch: Alleviating Overtrusting in Open-set Semi-supervised Learning
• Diffusion based Semantic Outlier Generation via Nuisance Awareness for Out-of-Distribution Detection
• Geometric Median (GM) Matching for Robust Data Pruning
• ConjNorm: Tractable Density Estimation for Out-of-Distribution Detection
• How Does Unlabeled Data Provably Help Out-of-Distribution Detection?
• How to Overcome Curse-of-Dimensionality for Out-of-Distribution Detection?
• Out-of-distribution Detection Learning with Unreliable Out-of-distribution Sources
• Learning to Augment Distributions for Out-of-Distribution Detection
• Diversified Outlier Exposure for Out-of-Distribution Detection via Informative Extrapolation
• SSB: Simple but Strong Baseline for Boosting Performance of Open-Set Semi-Supervised Learning
• Dream the Impossible: Outlier Imagination with Diffusion Models
• OpenOOD v1.5: Enhanced Benchmark for Out-of-Distribution Detection
• Feed Two Birds with One Scone: Exploiting Wild Data for Both Out-of-Distribution Generalization and Detection
• Is Fine-tuning Needed? Pre-trained Language Models Are Near Perfect for Out-of-Domain Detection
• Out-of-distribution Detection with Implicit Outlier Transformation
• Non-Parametric Outlier Synthesis
• Energy-based Out-of-Distribution Detection for Graph Neural Networks
• Watermarking for Out-of-distribution Detection
• Is Out-of-Distribution Detection Learnable?
• OpenOOD: Benchmarking Generalized Out-of-Distribution Detection
• Boosting Out-of-distribution Detection with Typical Features
• Out-of-Distribution Detection and Selective Generation for Conditional Language Models
• Extremely Simple Activation Shaping for Out-of-Distribution Detection
• Topological Structure Learning for Weakly-Supervised Out-of-Distribution Detection
• Domain Adaptation under Open Set Label Shift
• Dist-PU: Positive-Unlabeled Learning from a Label Distribution Perspective
• Positive Unlabeled Contrastive Learning
• Mitigating Neural Network Overconfidence with Logit Normalization
• Out-of-Distribution Detection with Deep Nearest Neighbors
• ViM: Out-Of-Distribution with Virtual-logit Matching
• Unknown-Aware Object Detection: Learning What You Don't Know from Videos in the Wild
• Training OOD Detectors in their Natural Habitats
• ReAct: Out-of-distribution Detection With Rectified Activations
• DICE: Leveraging Sparsification for Out-of-Distribution Detection
• Mixture Proportion Estimation and PU Learning: A Modern Approach
• Generalized Out-of-Distribution Detection: A Survey
• Generalized Out-of-Distribution Detection: A Survey
• On the Importance of Gradients for Detecting Distributional Shifts in the Wild
• Can multi-label classification networks know what they don't know?
• Semantically Coherent Out-of-Distribution Detection
• A Simple Fix to Mahalanobis Distance for Improving Near-OOD Detection
• Exploring the Limits of Out-of-Distribution Detection
• OodGAN: Generative Adversarial Network for Out-of-Domain Data Generation
• SSD: A Unified Framework for Self-Supervised Outlier Detection
• Energy-based Out-of-distribution Detection
• CSI: Novelty Detection via Contrastive Learning on Distributionally Shifted Instances
• Simple and Principled Uncertainty Estimation with Deterministic Deep Learning via Distance Awareness
• Proxy Anchor Loss for Deep Metric Learning
• Being Bayesian, Even Just a Bit, Fixes Overconfidence in ReLU Networks
• Positive-Unlabeled Reward Learning
• Towards neural networks that provably know when they don't know
• Reverse KL-Divergence Training of Prior Networks: Improved Uncertainty and Adversarial Robustness
• A Simple Baseline for Bayesian Uncertainty in Deep Learning
• SelectiveNet: A Deep Neural Network with an Integrated Reject Option
• Why ReLU networks yield high-confidence predictions far away from the training data and how to mitigate the problem
• Deep Anomaly Detection with Outlier Exposure
• A Simple Unified Framework for Detecting Out-of-Distribution Samples and Adversarial Attacks
• Predictive Uncertainty Estimation via Prior Networks
• Training Confidence-calibrated Classifiers for Detecting Out-of-Distribution Samples
• Enhancing The Reliability of Out-of-distribution Image Detection in Neural Networks
• Simple and Scalable Predictive Uncertainty Estimation using Deep Ensembles
• Wide Residual Networks
• Theoretical Comparisons of Positive-Unlabeled Learning against Positive-Negative Learning
• Towards Open Set Deep Networks
• LSUN: Construction of a Large-scale Image Dataset using Deep Learning with Humans in the Loop
• Dropout as a Bayesian Approximation: Representing Model Uncertainty in Deep Learning
• Deep Neural Networks are Easily Fooled: High Confidence Predictions for\n Unrecognizable Images
• PU Learning for Matrix Completion
• Describing Textures in the Wild

Cited by

• BootOOD: Self-Supervised Out-of-Distribution Detection via Synthetic Sample Exposure under Neural Collapse
• GOOD: Training-Free Guided Diffusion Sampling for Out-of-Distribution Detection

Related

• How Does Unlabeled Data Provably Help Out-of-Distribution Detection?
• Mining In-distribution Attributes in Outliers for Out-of-distribution Detection
• Training OOD Detectors in their Natural Habitats
• Out-Of-Distribution Detection with Diversification (Provably)
• When and How Does In-Distribution Label Help Out-of-Distribution Detection?