Our next STEM Bytes seminar is coming up on Friday, March 29 from 12 – 1pm PT on Zoom! To register, visit tinyurl.com/StemBytes3-29

𝐒𝐩𝐞𝐚𝐤𝐞𝐫: 𝐊𝐚𝐭𝐲 𝐅𝐞𝐥𝐤𝐧𝐞𝐫

🎤 Talk: ChatGPT, Are You Problematic?: Measuring and Mitigating Social Bias in Large Language Models

This talk will discuss how to define, measure, and mitigate social biases in AI systems like large language models, with a particular focus on homophobic and transphobic biases. We will focus on WinoQueer, a benchmark specifically designed to measure whether large language models (LLMs) encode biases that are harmful to the LGBTQ+ community. The benchmark was applied to several popular LLMs and find that off-the-shelf models generally do exhibit considerable anti-queer bias. Results show that LLM bias against a marginalized community can be somewhat mitigated by finetuning on data written about or by members of that community, and that social media text written by community members is more effective than news text written about the community by non-members. This method for community-in-the-loop benchmark development provides a blueprint for future researchers to develop community-driven, harms-grounded LLM benchmarks for other marginalized communities.

𝐒𝐩𝐞𝐚𝐤𝐞𝐫: 𝐉𝐢𝐚𝐪𝐢 𝐋𝐢𝐮

🎤 Talk: On Self-Similar Converging Shock Waves

The Guderley problem is a well-known hydrodynamic problem of a strong shock propagating radially in an ideal gas medium. The shock originates at infinity and collapses to a central point or an axis, where it is reflected and diverges toward infinity. In this talk, I will discuss the recent work on the rigorous proof of the construction of the Guderley solution. As a result, its solution is given in terms of similarity profiles as the Euler equations are transformed into a system of ordinary differential equations (ODEs). The proof is based on continuity arguments, nonlinear invariances, and barrier functions.