Open Source Models & Infra
Teaching AI to Improve AI: From Experimental Feedback to Meta-Evolution
Date, time, and room will be added once confirmed.
Talk overview
Large language models are evolving from coding assistants to research agents capable of proposing solutions, running experiments, and iteratively improving results. This talk discusses how execution-based experimentation can be transformed into reusable improvement capabilities, using the AI4AI loop and case studies from NatureBench. It also explores the potential for recursive self-improvement in AI systems.
