Skip to main content
Announcement

TrainDL-Team wins Best Poster Award at the INFOS 2023

With their poster on generative language models designed in the context of Grimm's fairy tales, researchers from the FU received the Best Poster Award at the INFOS 2023.

The 20th GI expert conference "Informatics and School" (INFOS) was organized by the University of Würzburg and took place this year under the motto "Informatics Education between Topicality and Timelessness". In addition to project coordinator Anna Lieckfeld, Till Zoppke and Viktoriya Olari from Freie Universität Berlin (FU) attended the conference on September 21 to present their poster in a one-minute pitch alongside 18 other scientific teams. Viktoriya Olari, Till Zoppke, and Ralf Romeike, who are responsible for designing and testing classroom activities in the TrainDL project, had created the poster "Language Models Unplugged with Grimm's Fairy Tales" for the occasion.

It presents a teaching task on generative artificial intelligence: Using nothing more than paper and pencil, students can design and test a simple language model. "Once upon a time..." Where can we go from there? What words can follow? With this exercise, students learn about training as well as testing a language model, learn more about the role of chance in generating new sentences, and also about the influence of the model on the sentences that can be generated.

In the context of the workshops conducted in the TrainDL project with computer science teachers and students, this exercise could already be tested. The participants found it suitable for teaching - even as early as grade five. The FU researchers were able to impress just as much at INFOS: their poster won the Best Poster Award 2023.

Curious? The poster is available for download here as a work-in-progress; the accompanying worksheets are currently still being revised.

If you are interested in the technical basics, the conception and the evaluation of the exercise, the paper accompanying the poster offers more insights: You can find it here.