AI Theory

This paper introduces Causally Fair DRO, a novel framework for robust optimization that addresses individual fairness concerns by incorporating causal structures and sensitive attributes, providing th…

This paper proves that Čech persistence diagrams converge to the true underlying shape precisely when using Wasserstein distances with p > m, where m is the submanifold dimension, significantly advanc…

This research introduces the first theoretical guarantees for warm-starting the celebrated Push-Relabel network flow algorithm, improving its speed using a predicted flow, while maintaining worst-case…

Unlocking the optimal variance estimation rate in compound decision theory under bounded means, this paper reveals a surprising (log log n/log n)² rate and introduces a rate-optimal cumulant-based est…

Researchers developed Spherical Convolutional Wasserstein Distance (SCWD) to more accurately validate climate models by considering spatial variability and local distributional differences.

AI explanations can be subtly manipulated to influence human decisions, highlighting the urgent need for more robust and ethical AI explanation design.

Noise in data surprisingly simplifies machine learning models, improving their interpretability without sacrificing accuracy; this paper quantifies this effect across various hypothesis spaces.

Recommendation systems must balance user satisfaction with fair item exposure. This research provides a theoretical model and empirical validation showing that user preference diversity can significan…

Recommender systems, when influenced by both users and creators, inevitably polarize; however, prioritizing efficiency through methods like top-k truncation can surprisingly enhance diversity.

Recommendation algorithms on UGC platforms face a critical trade-off: prioritizing user satisfaction reduces creator engagement, jeopardizing long-term content diversity. This research introduces a ga…

A new, nuisance-free Distributionally Robust Metric (DRM) is proposed for selecting robust Conditional Average Treatment Effect (CATE) estimators, improving the reliability of personalized decision-ma…

Self-attention, a key component of transformers, is revealed to be a projection of query vectors onto the principal components of the key matrix, derived from kernel PCA. This novel perspective leads…

Unrolled neural networks, trained via gradient descent, provably achieve optimal Bayesian inference for compressed sensing, surpassing prior-aware counterparts.

This paper reveals how large embedding dimensions and appropriate initialization guarantee convergence in Transformer training, highlighting Gaussian attention’s superior landscape over Softmax.

Adaptive gradient methods using AdaGrad stepsizes achieve optimal convergence rates for convex composite optimization problems, handling inexact oracles, acceleration, and variance reduction without n…

Poisson Private Representation (PPR) enables exact compression of any local differential privacy mechanism, achieving order-wise optimal trade-offs between communication, accuracy, and privacy.

This paper presents a novel framework for achieving tighter differential privacy guarantees via mechanism-specific amplification using subsampling.

Unified Covariate Adjustment (UCA) offers a scalable, doubly robust estimator for a wide array of causal estimands beyond standard methods.

Researchers unveil unelicitable backdoors in language models, using cryptographic transformer circuits, defying conventional detection methods and raising crucial AI safety concerns.

This work systematically investigates the approximation properties of Transformer networks for sequence modeling, revealing the distinct roles of key components (self-attention, positional encoding, f…