Mainz 2026 – wissenschaftliches Programm

Bereiche | Tage | Auswahl | Suche | Aktualisierungen | Downloads | Hilfe

Q: Fachverband Quantenoptik und Photonik

Q 67: Poster – Quantum Information

Q 67.23: Poster

Donnerstag, 5. März 2026, 17:00–19:00, Philo 2. OG

Evaluating the Impact of Expert-Curated vs. LLM-Generated Feedback on Novice Quantum Programmer Performance — •Lars Krupp^{1, 2}, Jonas Bley², Smriti Sharma^{1, 2}, Maximilian Kiefer-Emmanouilidis^{1, 2}, Paul Lukowicz^{1, 2}, and Jakob Karolus^{1, 2} — ¹DFKI Kaiserslautern — ²RPTU Kaiserslautern-Landau

The field of quantum computing (QC) faces a significant barrier to adoption due to the shortage of qualified educators. While online resources provide foundational knowledge, students often encounter coding challenges where a lack of timely, personalized assistance can severely stifle their learning progress. We present a user study on the PennyLane website using a plugin to inject different types of assistance into their error messages and investigate the efficacy of these assistance mechanisms for introductory QC coding tasks. Our study compares three types of assistance provided when a student’s code fails: standard error messages, expert-curated messages, and personalized, Large Language Model (LLM)-generated support. Unlike the non-personalized expert messages, the LLM-based system reads and interprets the student’s code and provides targeted guidance aimed at successful task completion. Learning gain is evaluated using a pre- and post-test design to assess the impact of these distinct support modalities on student understanding.

Keywords: Quantum Computing Education; Large Language Models; User Study