Researchers evaluated two large language models (LLMs) to analyze responses from digital simulations used in teacher education. The study compared DeBERTaV3 and Llama 3's performance in identifying characteristics in user interactions. Results showed that Llama 3 performed better than DeBERTaV3, especially when detecting new characteristics. This suggests that Llama 3 might be a more suitable choice for automatic evaluation of digital simulations where teacher educators need to introduce new characteristics. --- Why it matters: This study is important because it provides insights into the performance of large language models in analyzing user interactions in digital simulations. The findings can guide researchers and educators on choosing the most effective LLMs for automatic evaluation, which can help improve teacher education and training. Source: https://arxiv.org/abs/2407.20360