Optimizing generative AI by backpropagating language model feedback by Mert Yuksekgonul & Federico Bianchi & Joseph Boen & Sheng Liu & Pan Lu & Zhi Huang & Carlos Guestrin & James Zou ISBN 101038/S41586025086614 instant download

Optimizing generative AI by backpropagating language model feedback by Mert Yuksekgonul & Federico Bianchi & Joseph Boen & Sheng Liu & Pan Lu & Zhi Huang & Carlos Guestrin & James Zou ISBN 101038/S41586025086614 instant download

Recent breakthroughs in artifcial intelligence (AI) are increasingly driven by systems orchestrating multiple large language models (LLMs) and other specialized tools, such as search engines and simulators. So far, these systems are primarily handcrafted by domain experts and tweaked through heuristics rather than being automatically optimized, presenting a substantial challenge to accelerating progress. The development of artifcial neural networks faced a similar challenge until backpropagation and automatic diferentiation transformed the feld by making optimization turnkey. Analogously, here we introduce TextGrad, a versatile framework that performs optimization by backpropagating LLM-generated feedback to improve AI systems. By leveraging natural language feedback to critique and suggest improvements to any part of a system—from prompts to outputs such as molecules or treatment plans—TextGrad enables the automatic optimization of generative AI systems across diverse tasks. We demonstrate TextGrad’s generality and efectiveness through studies in solving PhD-level science problems, optimizing plans for radiotherapy treatments, designing molecules with specifc properties, coding, and optimizing agentic systems. TextGrad empowers scientists and engineers to easily develop impactful generative AI systems.