Corby Rosset

Corby Rosset is a Senior Researcher at Microsoft Research's AI Frontiers lab working on computer-use agents and LLM self-improvement. He is first author on Direct Nash Optimization (teaching language models to self-improve via general preferences) and a contributor to Microsoft's Fara-7B — an efficient open-weight agentic small model for computer use that scored 73.5% on WebVoyager — and its successor Fara1.5. His session offers a researcher's view from the team building open-weight browser-automation models.