Computer Science PGR Seminar Tilcia Woodville-Price & Zipei Li

All are welcome to listen to our speakers Tilcia Woodville-Price and Zipei Li.

Tilcia Woodville-Price will present Visualizing Uncertainty with Icon Arrays: Communicating the Diagnostic Accuracy of a Cancer Screening Test

Icon arrays are widely used in medical risk communication to translate complex numerical information into simpler visual representations. Uncertainty, however, is rarely disclosed to non-expert audiences despite its pervasiveness in scientific and visualization research alike. This talk will present the results from a recent experiment which sought to examine how visualizing uncertainty impacts non-expert audiences' perspectives by incorporating visualization techniques that encode uncertainty into icon arrays.

Tilcia is a PhD student co-supervised by Computer Science and Medicine. Her main research area is medical information visualisation, with an emphasis on uncertainty visualisation, and data-driven storytelling.

Zipei Li will present What is really difficult for LLM planners?

Large Language Models (LLMs) have shown promising performance on planning tasks, yet their failures reveal important structural limitations. In this talk, I revisit our previous empirical study and examine what is fundamentally difficult for LLM planners.
First, through a relaxation-based evaluation, we show that many plans fail strict symbolic validation due to poor grounding into primitive actions rather than wrong high-level intent, highlighting difficulties in rule obedience. Second, we analyse how plan generation style affects performance: adding explicit state-based chain-of-thought improves action executability but can harm goal achievement, exposing a trade-off between local consistency and long-horizon reasoning. Finally, we explore traditional complexity metrics such as optimal plan length and object number, presenting preliminary evidence that LLM performance may depend more on structural properties of the planning problem.
These findings offer a deeper perspective on the challenges of LLM-based planning and motivate more structured and hybrid approaches.

Zipei is a second-year PhD student working on LLM-based robotic planning.