226 future engineers are learning to work with LLMs by controlling a Spot robot using prompts

A gamified workshop involves students writing prompts to plan robot paths: 81.5 per cent of students say they learnt a great deal, and the way they phrase their prompts varies depending on the model.

How do you teach engineering students to collaborate with a large language model in a critical way? Researchers designed a fun and repeatable workshop on a large scale, followed by 226 master's students over the course of ten sessions. From their phones, participants wrote prompts to have a classic LLM and then a reasoning model solve increasingly complex grid navigation problems.

The system returned a plan that could be executed by a robot, a visualization of the trajectory and an automatic score, before a final live demonstration on a Boston Dynamics Spot robot. In the end-of-workshop questionnaire, 81.5% of students said they had learned a lot and 91% said they were very engaged.

The analysis of the prompts shows an evolution of strategies: step-by-step instructions for the model without reasoning, then higher-level instructions for the complex problems entrusted to the reasoning model. The authors consider this transition from simulation to reality suitable for teaching the verification and collaboration necessary for a responsible use of LLMs in engineering; the workshop code is published in open access.