Researchers have developed a fully automated framework that uses large language models to design and optimize neural networks for handwritten text recognition. The framework, called LLM-Driven AutoML, uses three different language models (GPT-5, GPT-4o, and Claude Sonnet 4) to generate, train, and refine neural network architectures without human intervention. The framework was tested on datasets in Arabic, Persian, and English and achieved high accuracy rates, with some models reaching up to 98.1% accuracy. This approach could enable scalable and reproducible handwriting recognition across languages. --- Why it matters: This work matters because it shows that large language models can be used as effective agents for neural architecture search, which is a key challenge in machine learning research. The ability to automate the design of neural networks could lead to significant improvements in areas like handwritten text recognition, where current methods often require manual tuning and are not easily scalable. Source: https://arxiv.org/abs/2607.15509