Lior Wolf

Examining methods to enhance object localization in weakly supervised and unsupervised settings.

A new method reduces parameters in transformer models while maintaining performance.

This study examines how language models learn and store information during training.

A new approach enhances multi-subject image generation using layout manipulation.

New methods combine linear block codes and neural networks for better data transmission.

Mamba’s context-extension method improves handling of long sequences without additional training.

A new method enhances accuracy in counting objects in generated images.

Revolutionizing how we perceive the world in 360 degrees.

A new method improves 3D image edits using attention warping for better consistency.

Discover how Reversed Attention improves language models' learning and decision-making.

Learn how MAS boosts language model performance in chatbots and reasoning tasks.