Description
TextGrad is a powerful framework building automatic ``differentiation'' via text. TextGrad implements backpropagation through text feedback provided by LLMs, strongly building on the gradient metaphor
Papers Using This Method
Gen-n-Val: Agentic Image Data Generation and Validation2025-06-05metaTextGrad: Automatically optimizing language model optimizers2025-05-24Can Textual Gradient Work in Federated Learning?2025-02-27How to Correctly do Semantic Backpropagation on Language-based Agentic Systems2024-12-04TextGrad: Automatic "Differentiation" via Text2024-06-11