Luke Zettlemoyer
University of Washington professor and Senior Research Director at Meta; co-author of RoBERTa and QLoRA
- Professor, Paul G. Allen School of Computer Science & Engineering, University of Washington12
- Senior Research Director, Meta FAIR, Meta1
Luke Zettlemoyer is a Professor in the Paul G. Allen School of Computer Science & Engineering at the University of Washington and a Senior Research Director at Meta FAIR. The Allen School's faculty page describes research on empirical methods for natural language semantics, new tasks and datasets, and how to train and deploy large language models, and notes that Zettlemoyer is a Fellow of the Association for Computational Linguistics. Zettlemoyer co-authored the RoBERTa pretraining study, the deep contextualized word representations paper, and QLoRA, a finetuning method that backpropagates through a frozen 4-bit quantized model into low-rank adapters so that a 65B parameter model can be finetuned on a single 48GB GPU.12453
Work
QLoRA: Efficient Finetuning of Quantized LLMs (external site: arxiv.org)Published research
Co-author with Tim Dettmers, Artidoro Pagnoni, and Ari Holtzman of the QLoRA paper, which introduces 4-bit NormalFloat, double quantization, and paged optimizers to cut finetuning memory.3
RoBERTa: A Robustly Optimized BERT Pretraining Approach (external site: arxiv.org)Published research
Listed among the authors of the 2019 RoBERTa paper, a replication study of BERT pretraining.4
Deep contextualized word representations (external site: arxiv.org)Published research
Co-author of the 2018 paper introducing word vectors learned from the internal states of a deep bidirectional language model.5
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.