Yao Li

A new method enhances targeted attacks using easy samples in neural networks.

A new benchmark assesses language models on scientific coding challenges across multiple fields.

TART boosts model accuracy while defending against adversarial attacks.

Researchers study how jets behave in high-energy nuclear collisions and their interactions.

Algorithms confront bias and delays in user preference feedback.

Investigating how heavy-flavor jets behave in high-energy nuclear collisions.

AdaScale automates deep learning for better mobile performance and privacy.

Denoising models face challenges from adversarial noise but new strategies offer hope.