This paper establishes a large deviation principle for the generalization error of interpolating classifiers in the overparametrized regime.

This paper accelerates conformal prediction by incorporating approximate leave-one-out estimators and establishes asymptotic coverage and efficiency.

This paper develops statistical learning theory for gradient boosting in Peaks-over-Threshold modeling using Generalized Pareto distributions, deriving error bounds and reducing gradient correlation.

This paper benchmarks and improves conformal predictions for image classification using vision-language models by proposing a simple non-linear transformation of cosine similarities.

This paper presents a method to equip existing motion planning algorithms with probabilistic task-completion guarantees on systems with unknown dynamics using a planner-agnostic constraint-tightening…

The paper demonstrates that fine-tuning safety guard models on benign data can catastrophically collapse their safety alignment, proposing Fisher-Weighted Safety Subspace Regularization (FW-SSR) to ac…

This paper introduces Conformal Risk Sharing, a framework to fairly distribute the financial burden of rare adverse events while guaranteeing that no individual is made worse off.

The paper establishes tight upper and lower bounds on the statistical cost of approximate machine unlearning for smooth strongly convex losses, showing that the optimal unlearning rate depends critica…

The paper introduces Fine-Tuning Integrity (FTI), a security goal that uses Succinct Model Difference Proofs (SMDPs) to cryptographically prove that a fine-tuned model update adheres to specific struc…

The paper introduces Conformal Generation (Conf-Gen), a novel framework that adapts conformal risk control to provide formal uncertainty guarantees for unsupervised generative models like LLMs and ima…

The paper formalizes the problem of representation identifiability in supervised learning, showing that a representation property is identifiable if and only if it is constant across all possible fact…

This paper proposes a simple real-time monitor for LLMs that turns an external verifier signal into an alarm decision by thresholding, showing competitiveness with advanced monitors in mathematical re…

The paper introduces PRAXIS, a novel algorithm that efficiently approximates the computation of 'Rashomon sets' for decision trees, significantly reducing memory and runtime complexity.

A new boosting algorithm that strong learns concept classes closed under O(log 1/γ)-XOR using O(log 1/ε) calls to a γ-advantage weak learner and additional samples, by connecting boosting with list-de…

The paper introduces SafetyDrift, a predictive model that forecasts when AI agents will violate safety protocols by analyzing the cumulative risk across sequences of individually safe actions.

The paper conducts a preliminary safety evaluation of the open-weight LLM Kimi K2.5, finding that while it is highly capable, it exhibits concerning dual-use risks, particularly regarding CBRNE misuse…

The paper introduces Nested Contextual Causal Bandits (NCCBs) to model multi-timescale sequential decisions and proposes a certified policy optimization method, NCTS, that provides quantifiable risk b…