Tat-Seng Chua

Setokim enhances the fusion of visual and text understanding through innovative tokenization.

Combining image generation and retrieval for better visual information access.

Softmax-DPO introduces negative samples for better user preference alignment in recommendations.

DisMAE enhances model generalization across domains using unlabeled data.

Combining images and text enhances predictions of future events.

Examining the strengths and weaknesses of VideoQA systems in understanding video content.

A new approach enhances video question answering through scene text recognition.

A new approach enhances malware detection while resisting adversarial attacks.

AI learns to create art through self-feedback for better image alignment.

Discover the knowledge boundaries of LLMs and their challenges.