Shixiang Shane Gu
Senior staff research scientist at Google DeepMind on Gemini; co-author of Gumbel-Softmax and zero-shot CoT
- Senior Staff Research Scientist (Gemini Thinking team), Google DeepMind1
Shixiang Shane Gu is a Senior Staff Research Scientist at Google DeepMind, focusing on the Gemini Thinking team; Gu previously led the multilinguality team for Gemini post-training. Before that, Gu was a senior researcher on the ChatGPT team at OpenAI, and earlier a founding member of the Google Brain Robotics team. Gu co-authored "Categorical Reparameterization with Gumbel-Softmax" (2016), which introduced a differentiable gradient estimator for categorical variables, and "Large Language Models are Zero-Shot Reasoners" (2022), on eliciting step-by-step reasoning without task-specific examples.123
Work
Gumbel-Softmax (external site: arxiv.org)Published research
Co-author, with Eric Jang and Ben Poole, of the 2016 paper presenting a gradient estimator that replaces non-differentiable categorical samples with samples from a Gumbel-Softmax distribution.2
Large Language Models are Zero-Shot Reasoners (external site: arxiv.org)Published research
Co-author of the 2022 paper on zero-shot chain-of-thought prompting of large language models.3
Other documented roles
Roles the sources document but that are not confirmed as current.
- Senior researcher on the ChatGPT team, OpenAI1
- Founding member of the Google Brain Robotics team, Google Brain1
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.