Learning Theory

Attention layers provably solve single-location regression ICLR 2025 A Two-Stage Learning-to-Defer Approach for Multi-Task Learning ICML 2025 A Unified Framework for Generalization Error Analysis of Learning with Arbitrary Discrete Weak Features ICML 2025 A Unified Interpretation of Training-Time Out-of-Distribution Detection ICCV 2025 A Unified Theoretical Analysis of Private and Robust Offline Alignment: from RLHF to DPO ICML 2025 A Unified Theory of Supervised Online Learnability ALT 2025 AutoEval Done Right: Using Synthetic Data for Model Evaluation ICML 2025 Automatic debiasing of neural networks via moment-constrained learning CLeaR 2025 A Variational Information Theoretic Approach to Out-of-Distribution Detection ICML 2025 Average Sensitivity of Hierarchical k-Median Clustering ICML 2025 Axioms for AI Alignment from Human Feedback AAAI 2025 Backdoor Attacks in Token Selection of Attention Mechanism ICML 2025 Balancing the Scales: A Theoretical and Algorithmic Framework for Learning from Imbalanced Data ICML 2025 Batch Ensemble for Variance Dependent Regret in Stochastic Bandits AAAI 2025 BaxBench: Can LLMs Generate Correct and Secure Backends? ICML 2025 BECAME: Bayesian Continual Learning with Adaptive Model Merging ICML 2025 Behavioural vs. Representational Systematicity in End-to-End Models: An Opinionated Survey ACL 2025 Bellman Unbiasedness: Toward Provably Efficient Distributional Reinforcement Learning with General Value Function Approximation ICML 2025 Benchmarking Deep Search over Heterogeneous Enterprise Data EMNLP 2025 Benchmarking Large Language Models for Cryptanalysis and Side-Channel Vulnerabilities EMNLP 2025 Benchmarking Quantum Reinforcement Learning ICML 2025 Benchmark Profiling: Mechanistic Diagnosis of LLM Benchmarks EMNLP 2025 Benefits of Early Stopping in Gradient Descent for Overparameterized Logistic Regression ICML 2025 Benign Overfitting in Out-of-Distribution Generalization of Linear Models ICLR 2025 Benign Overfitting in Token Selection of Attention Mechanism ICML 2025