Philip Torr

A new system for assessing language models using real-world data streams.

This article discusses a new benchmark for combining images and text to find events in videos.

CRAB enhances testing for language models in real-world environments.

VQA models may expose private information despite advanced techniques.

A new method enhances image editing through clearer instruction handling.

This article discusses safety issues in text-to-image models and proposes solutions.

A new method for robots to navigate effectively without extensive training.

A project providing artwork descriptions in 28 languages for broader engagement.

A look into shape recognition challenges for machines and the way forward.

This method simplifies the creation of match-cuts for filmmakers of all levels.

A new method for efficiently pruning image-generating AI models while preserving quality.

A new method enhances video generation by applying motion from one video to another.