Bill Howe

Examining when LLMs should refrain from answering questions.

Examining smaller AI models' effectiveness and cost compared to larger counterparts.

A new benchmark enhances evaluation of pedestrian pathways for better urban accessibility.

Examining how language models can refuse to answer for improved safety.

Using vision-language models to improve urban mapping accuracy and accessibility.

ML-EAT offers a structured approach to uncovering bias in language technologies.

Examining how AI learns human-like biases from facial impressions.

A look at synthetic data and its role in privacy.