Date: November 20, 2026
Speaker: Nathan Lambert, Machine Learning Researcher
Title: Making frontier post-training open and scientific
Host: Claire Cardie

 A color photo of a man.

Abstract: In this talk, I reflect on what it takes to do academic and open research that is not only relevant at the rapidly evolving AI frontier, but also impacts it. This will include reflections on what has worked in my past work building open language models (Olmos), and how I’m designing new problems in a more agentic era of research. This talk will cover foundations of post-training and early work in my more recent areas of focus, including character training, agentic RL, and measuring the automation of pieces of AI research. The talk will conclude with principles for choosing research problems as an academic and some ideas that I am interested in collaborating on.

Bio: Nathan Lambert is a co-founder and Executive Director of Trillium Labs, a non-profit with the mission of fostering the open science of frontier AI. He also writes Interconnects AI, a well read technical newsletter on open models and AI research. He was recently a Senior Research Scientist and post-training lead at the Allen Institute for AI, where he co-led the team building open language models (Olmo 1, 2, and 3; Tülu 2 and 3). Previously, he helped build an RLHF research team at HuggingFace and received his PhD from the University of California, Berkeley.