Mohan Kankanhalli

A new approach aligns language models with video content using textual simulations.

Unlearning enhances privacy in recommender systems while maintaining recommendation quality.

SCAN method improves machine learning by using less data efficiently.

Examining the vulnerabilities and defenses of new AI models.

A new approach tackles biases in image-text models effectively.

VidHal benchmarks video models' ability to accurately interpret content.

Researchers uncover vulnerabilities in Multi-Modal Large Language Models through clever tactics.