Projects

Please see the Publications page for specific project developments.

Identifying and Mitigating Bias in LLMs

How can we measure or eliminate social, political, and cognitive biases in LLMs to ensure predictable and robust use?
Figure showing the dataset creation process for bias identification.

Information Extraction for Grounded Educational Games

How can we process educational content (e.g., textbooks) to create high-quality assets for educational games?
Diagram of GPT-2 finetuning for improved image generation. Example takes the image subject, London, and generates the prompt \

Semantic Parsing with Language Models

How best can we use the capabilities of LLMs to improve semantic parsing systems that require precise meaning representations?
Minimalistic logo for unscoped episodic logical forms.