IIDS AI Faculty Fellowship Presentations

April 24, 2025 12:30pm – 1:30pm
IRIC Atrium

Dilshani Sarathchandra
Associate Professor
Department of Culture, Society & Justice

“Generative AI for Conversing with Qualitative Data”
This project examines the application of multimodal Generative AI models for sociological research. We present a preliminary benchmark for evaluating Generative AI models for image-based data. By applying the benchmark criteria for a qualitative project on #scoliosis, we identify some important considerations for image-to-text sociological research including structured output generation and managing prompt sensitivity. 

Jennifer Johnson-Leung
Professor
Department of Mathematics and Statistical Science

“Exploring the Potential and Limits of LLMs as Mathematical Research Assistants”

Generative AI tools based on large language models are increasingly capable of engaging with complex ideas in current mathematical research. Yet, significant challenges remain when it comes to using these tools effectively in the daily work of mathematicians. In this talk, I’ll explore both the strengths and limitations of these models, drawing on my experience this semester using AI to assist in refereeing a research article. I’ll discuss experiments with both locally deployed and publicly available models, offering a candid look at where these tools shine—and where they still fall short.

Similar Posts

Reasoning Models in Generative AI: How the Next Generation of LLMs Can Think

Reasoning Models in Generative AI: How the Next Generation of LLMs Can Think

The latest frontier “thinking models” can apparently begin to match the reasoning performance of humans. We will do a technical deep dive on the likely underpinnings of the latest generation of frontier reasoning models, including OpenAI’s o1 and o3. We will discuss and implement in Python a simple LLM that uses self-taught reasoning (STR) and Q-learning at both training and inference.