Researchers have developed a method using transformers to improve artificial intelligence systems' ability to perform analogical reasoning, a key aspect of human intelligence. They trained models on letter-string analogies by including copy tasks in the training data, which helped them attend to the most informative elements and generalize to new alphabets. The approach also enabled some generalization to novel transformations, although not completely novel ones. The study found that the model's performance was comparable to state-of-the-art models on the task, and the researchers were able to identify an algorithm for solving one analogy task by analyzing the model's computations. --- Why it matters: This research matters because it addresses a long-standing challenge in AI: developing systems capable of robust analogical reasoning. Improving this ability could lead to more effective problem-solving and decision-making in various applications, such as natural language processing and computer vision. Source: https://arxiv.org/abs/2604.06501