🏢 University of Pennsylvania

Researchers unveil a novel framework to dissect bias in large-scale visual datasets, identifying unique visual attributes and leveraging language models for detailed analysis, paving the way for creat…

Boosting hippocampal spatial resolution surprisingly shrinks its contextual memory capacity, revealing a crucial trade-off between precision and context storage.

This paper introduces efficient algorithms for learning under arbitrary covariate shift, addressing limitations of prior approaches by enabling classifiers to abstain from predictions in high-shift sc…

Networks trained on continuous sensory data spontaneously develop place cell-like responses, demonstrating that time-encoded experience can create spatial maps in the brain.

Boost RL performance in large state spaces by efficiently learning a policy competitive with the best combination of existing base policies!

One-shot dualization aligns large language models with safety constraints efficiently, eliminating iterative primal-dual methods for improved stability and reduced computational burden.

Conformal Prediction with Length Optimization (CPL) achieves shorter, conditionally valid prediction sets by optimizing length while ensuring coverage across various covariate shifts.

Boost equivariant model training by strategically relaxing constraints during training, enhancing optimization and generalization!

Hybrid RL algorithms achieve sharper error/regret bounds than existing offline/online RL methods in linear MDPs, improving sample efficiency without stringent assumptions on behavior policy quality.

Generative adversarial model-based optimization via adaptive source critic regularization (aSCR) significantly boosts offline optimization accuracy.

Fairness-aware estimation of graphical models (GMs) tackles bias in GM estimations by integrating graph disparity error and a tailored loss function into multi-objective optimization, effectively miti…

ISED: a novel, data-efficient algorithm learns neural programs by sampling from neural predictions to estimate gradients of black-box components, outperforming baselines on various benchmarks.

Constrained diffusion models, trained via a novel dual approach, achieve optimal trade-offs between data fidelity and user-defined distribution constraints, enabling fairer and more controlled data ge…

AR-Pro uses generative models to create counterfactual explanations for anomaly detection, formally specifying what a non-anomalous version should look like and improving interpretability.

Approximation-Aware Bayesian Optimization (AABO) boosts high-dimensional Bayesian optimization by jointly optimizing model approximation and data acquisition, achieving superior efficiency and perform…

Acoustic Volume Rendering (AVR) revolutionizes realistic audio synthesis by adapting volume rendering to model acoustic impulse responses, achieving state-of-the-art performance in novel pose synthesi…

KnoBo enhances deep learning models for medical image analysis by incorporating knowledge priors from medical textbooks, boosting out-of-domain performance by up to 32.4%.