A study from Universitat Politècnica de València found that newer large language models (LLMs) like BLOOM, Llama, and GPT are less likely to admit uncertainty in their answers, making them less reliable. While they perform better on complex problems, they often guess on simpler questions, leading to incorrect responses. The study highlights a lack of improvement in basic problem-solving despite advancements in technology. Teljes cikk (Euronews.com)