Open Source Models & Infra
Building Self-Improving and Long-Horizon Agents
Date, time, and room will be added once confirmed.
Talk overview
Building Self-Improving and Long-Horizon Agents explores how large language models can autonomously enhance their operational systems. The talk introduces Self Harness, where an LLM acts as its own harness engineer, and Harness of Harness, enabling continuous improvement over extended periods. It also touches on applying these methods to theoretical computer science for AI-assisted research.
