UCR AI Research Brown Bag Presentation Series

Join us for a low-stakes, supportive space to support and engage with UCR graduate students’ research progress across the UCR community. This event is hosted by the UCR Library, RAISE@UCR, GradSuccess, and the XCITE Center for Teaching and Learning.

This is an informal seminar series spotlighting AI research from UCR graduate students. This is a great opportunity to engage with emerging scholarship, ask questions, and connect with researchers pushing the boundaries of artificial intelligence.

Pizza and drinks will be provided while supplies last! No need to bring your own lunch/brown bag.

Presentations:

Just Do It!? Computer-Use Agents Exhibit Blind Goal-Directedness - A Safety Study on Computer-Use Agents with Microsoft and NVIDIA

Erfan Shayegani, PhD student, Computer Science

In this talk Erfan will talk about adversarial attacks on multi-modal language models. Computer-Use Agents (CUAs) are an increasingly deployed class of agents that take actions on GUIs to accomplish user goals. In this paper, we show that CUAs consistently exhibit Blind Goal-Directedness: a bias to pursue goals regardless of feasibility, safety, reliability, or context. The paper has been accepted to ICLR 2026 with Microsoft Research.

From JWST to Roman: A Physics-Informed AI Super-Resolution Framework for Grism Spectroscopy

Aryana Haghjoo, Ph.D. student, Astronomy

The Roman Space Telescope's High Latitude Wide Area Survey will deliver grism spectroscopy at R~460–900 for roughly ten million emission-line galaxies, but at this resolution key diagnostic features introduce systematic biases in redshift measurements and derived physical quantities. In this talk, Aryana introduces a three-stage physics-informed deep-learning framework for spectral super-resolution, trained on 1,187 paired JWST/NIRSpec prism–grating observations from JADES, that enhances low-resolution galaxy spectra by a factor of 10 in resolving power (R~100 to R~1000).

All UCR community members are welcome to attend. Make sure to register with your UCR email. Zoom link will be emailed to registered participants one hour before the workshop starts. Participants are expected to follow and uphold UC Riverside's Principles of Community.