Stanford CS329A Self-Improving AI Agents | Part 4 | Learning from Feedback with Tools/Code

Want to dive deeper? This curriculum is covered in the following online courses: - Agentic AI professional education course: https://learn.stanford.edu/agentic-ai-2026.html - XCS329 graduate course: https://online.stanford.edu/courses/cs329a-self-improving-ai-agents A similar curriculum is covered in XCS329z https://online.stanford.edu/courses/cs329z-engineering-ai-agents Follow along with the course schedule and syllabus: https://cs329a.stanford.edu/ Aakanksha Chowdhery Adjunct Professor of Computer Science, Stanford University View the course playlist: https://www.youtube.com/playlist?list=PLangBM27OtEA Video Summary: This lecture recording from Stanford's CS329A, Self-Improving AI Agents, taught by Aakanksha Chowdhery on October 3, 2025, covers three approaches to improving language models through feedback. ReAct interleaves chain-of-thought reasoning with tool-calling actions and is evaluated on HotpotQA, FEVER, and WebShop. RLEF, or Reinforcement Learning from Execution Feedback, trains coding agents using public and private unit test results within a PPO training loop, evaluated on CodeContests. Constitutional AI, developed by Anthropic, uses a written set of principles along with model self-critique and revision to train a preference model through reinforcement learning from AI feedback rather than human feedback. The lecture compares how each method sources its feedback signal, from environment interaction to execution results to AI-generated critique, and reviews related work including WebGPT, Code Monkeys, and SWE-bench. Speaker Bio: Aakanksha Chowdhery Adjunct Professor of Computer Science, Stanford University Dr. Aakanksha Chowdhery is pushing the frontier of agentic LLMs, focusing on recursive self-improvement and long-horizon agents that learn and deploy in the real world. She is one of the few researchers globally who has led frontier model training end-to-end, across both dense and mixture-of-experts (MoE) architectures. At Google, she led the 540B PaLM model, the largest densely trained language model in the world at the time. She subsequently drove pre-training and scaling of Gemini's MoE models across multiple generations, and contributed key components to PaLM-E, Med-PaLM, and the Pathways infrastructure underpinning Google's large-model efforts. She went on to build and lead pretraining teams for open intelligence efforts at Reflection and Meta. Earlier, she held research roles at Microsoft Research and Princeton. At Stanford, where she earned her PhD, she teaches CS329A (Self-Improving AI Agents) and serves as Program Chair for MLSys 2026.