Deceptive Risk Minimization: Out-of-Distribution Generalization by Deceiving Distribution Shift Detectors

2025/09/15 by Anirudha Majumdar, Majumdar, Anirudha
Computer Science · Engineering · #Artificial Intelligence (cs.AI) #Distributed Sensor Networks and Detection Algorithms #FOS: Computer and information sciences #Fault Detection and Control Systems #Machine Learning (cs.LG) #Robotics (cs.RO)

paper · pdf · doi:10.48550/arxiv.2509.12081

openalex publication_date 2025/09/15 · openalex created_date 2025/10/12 · openalex updated_date 2026/07/28

Abstract

This paper proposes deception as a mechanism for out-of-distribution (OOD) generalization: by learning data representations that make training data appear independent and identically distributed (iid) to an observer, we can identify stable features that eliminate spurious correlations and generalize to unseen domains. We refer to this principle as deceptive risk minimization (DRM) and instantiate it with a practical differentiable objective that simultaneously learns features that eliminate distribution shifts from the perspective of a detector based on conformal martingales while minimizing a task-specific loss. In contrast to domain adaptation or prior invariant representation learning methods, DRM does not require access to test data or a partitioning of training data into a finite number of data-generating domains. We demonstrate the efficacy of DRM on numerical experiments with concept shift and a simulated imitation learning setting with covariate shift in environments that a robot is deployed in.

Citations

• π_0.5: a Vision-Language-Action Model with Open-World Generalization
• Invariant Learning with Annotation-free Environments
• Probing out-of-distribution generalization in machine learning for materials
• Testing exchangeability by pairwise betting
• A Survey on Efficient Training of Transformers
• A System-Level View on Out-of-Distribution Data in Robotics
• Online Distribution Shift Detection via Recency Prediction
• Continual Test-Time Domain Adaptation
• E-detectors: a nonparametric framework for sequential change detection
• Learning Optimal Conformal Classifiers
• MEMO: Test Time Robustness via Adaptation and Augmentation
• Towards Out-Of-Distribution Generalization: A Survey
• Task-Driven Detection of Distribution Shifts with Statistical Guarantees for Robot Learning
• Toward Causal Representation Learning
• Generalizing to Unseen Domains: A Survey on Domain Generalization
• Nonlinear Invariant Risk Minimization: A Causal Approach
• Testing exchangeability: fork-convexity, supermartingales, and\n e-processes
• Transporter Networks: Rearranging the Visual World for Robotic Manipulation
• The Risks of Invariant Risk Minimization
• Online Meta-Learning for Multi-Source and Semi-Supervised Domain Adaptation
• Out-of-Distribution Generalization via Risk Extrapolation (REx)
• Fast Differentiable Sorting and Ranking
• Invariant Risk Minimization Games
• Invariant Risk Minimization
• Differentiable Ranks and Sorting using Optimal Transport
• Failing Loudly: An Empirical Study of Methods for Detecting Dataset Shift
• Detecting and Correcting for Label Shift with Black Box Predictors
• Deep Residual Learning for Image Recognition
• Causal inference using invariant prediction: identification and confidence intervals
• Optimal Transport for Domain Adaptation
• Unsupervised Domain Adaptation by Backpropagation
• A Reduction of Imitation Learning and Structured Prediction to No-Regret\n Online Learning
• Gemini Robotics: Bringing AI into the Physical World

Related

• On the Connection between Invariant Learning and Adversarial Training for Out-of-Distribution Generalization
• Task-Driven Detection of Distribution Shifts with Statistical Guarantees for Robot Learning
• Out-of-Distribution Generalization via Risk Extrapolation (REx)
• Transformation-Invariant Learning and Theoretical Guarantees for OOD Generalization
• Learning Representations that Support Robust Transfer of Predictors