LLMs and clustering methods assessed for team problem-solving exercises

A new research paper proposes and evaluates two methods for assessing student teams in tabletop exercises (TTXs), which are used to prepare learners for workplace tasks and crisis responses. The study compared a clustering approach, which groups teams with similar task approaches for faster feedback, against the use of large language models (LLMs) for assessment. While clustering proved valid and reliable, LLM assessment showed varying results, with GPT-5.2 demonstrating lower error rates than GPT-4o when compared to instructor scores. These methods have been integrated into the open-source INJECT TTX learning platform. AI

IMPACT Introduces novel LLM-based assessment techniques for educational exercises, potentially improving feedback efficiency and accuracy.

RANK_REASON Academic paper presenting new methods for assessment in computing education. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.AI →

• GPT-4o
• GPT-5.2
• Valdemar Švábenský

• paper
• product

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →