Collin Burns
AI researcher formerly at OpenAI and Anthropic; first author of weak-to-strong generalization research
Collin Burns is an AI researcher who previously worked at OpenAI and Anthropic. Burns is first author of "Weak-to-Strong Generalization", which studies whether supervision from a weaker model can elicit the full capabilities of a much stronger one, as an analogy for humans supervising future superhuman models. Burns was a PhD student at UC Berkeley co-advised by Jacob Steinhardt and Dan Klein, and is also co-first author of "Discovering Latent Knowledge in Language Models Without Supervision" and a co-author of the MMLU benchmark paper.12345
Work
Weak-to-Strong Generalization: Eliciting Strong Capabilities With Weak Supervision (external site: arxiv.org)Published research
First-listed author and one of the paper's primary contributors; the 2023 study tests whether weak model supervision can elicit the full capabilities of a much stronger pretrained language model.31
Discovering Latent Knowledge in Language Models Without Supervision (external site: arxiv.org)Published research
Co-first author of the ICLR 2023 paper, which introduces an unsupervised method for answering yes-no questions from a language model's internal activations alone.41
Measuring Massive Multitask Language Understanding (MMLU) (external site: arxiv.org)Published research
Co-author of the 2020 paper proposing a test of a text model's multitask accuracy across 57 tasks.51
Earlier roles
This profile covers public professional work only, from the sources below. It is not an endorsement or a ranking. To correct or remove it, use the correction form.