AI grades student answers

A team from the Language and AI in Education lab, led by Prof. Dr. Detmar Meurers, took first place in the BEA 2026 Shared Task on Rubric-based Short Answer Scoring for German.

The competition took place as part of the 21st workshop “Innovative Use of NLP for Building Educational Applications (BEA 2026)” in San Diego (USA). The workshop is part of the annual conference of the Association for Computational Linguistics (ACL), the world’s leading scientific society for language processing and artificial intelligence.

The challenge of the Shared Task was to train AI models so that they can reliably and accurately assess short answers from pupils in STEM subjects, based on detailed assessment rubrics. In future, such methods could, for example, support teachers in their marking or enable adaptive feedback in digital learning environments.

The IWM team’s AI approach prevailed against strong international competition, securing first place in both the ‘unknown answers’ and ‘unknown questions’ categories.

‘This success illustrates that reliable, automated assessments of free-text responses are only possible through specific approaches, not simply by using general AI applications,’ emphasises Detmar Meurers. ‘If such developments are supported in Germany, these methods can assist teachers in the long term and enable personalised feedback in digital learning environments.’

The paper was published by the ACL and awarded the prize for the best paper in the Shared Task:

Belcher, K., De Kuthy Meurers, M., De Kuthy, K. & Meurers, D. (2026). IWM-DKM at BEA 2026 Shared Task 2: Supplementing Supervised Fine-Tuning for Rubric-Based Short Answer Scoring. https://doi.org/10.18653/v1/2026.bea-1.90