Scholar
DJ Strouse
Google Scholar ID: K8E0T7MAAAAJ
Research Scientist, DeepMind
Citations & Impact
Citations
5,416
H-index
15
i10-index
17
Publications
20
Co-authors
0
Publications
Browse publications on Google Scholar (top-right) ↗
Resume (English only)
Academic Achievements
- Gemini 1.5: Unlocking multimodal understanding across millions of tokens of context (2024); Tokenization counts: the impact of tokenization on arithmetic in frontier LLMs (2024); Confronting reward model overoptimization with constrained RLHF (ICLR 2024 Spotlight).
Research Experience
- Member of Technical Staff at OpenAI, working on large-scale reinforcement learning for reasoning; Research Scientist at DeepMind in New York, working on improving the reasoning capabilities of frontier models with the Blueshift team; Interned at DeepMind (with Matt Botvinick), Stanford University (with Kwabena Boahen), Institute for Quantum Computing (with Andrew Childs), and Spotify NYC with their machine learning team.
Education
- PhD in Physics from Princeton University, advised by David Schwab and Bill Bialek; Master's from the University of Cambridge, advised by Mate Lengyel, as a Churchill Scholar; Bachelor's in Physics and Mathematics from the University of Southern California, worked with Bartlett Mel and Paolo Zanardi.
Background
- Broadly interested in improving reasoning capabilities in frontier models. Previously, worked on various topics in reinforcement learning, including exploration and training agents to play cooperative games with humans. PhD thesis focused on applications of the information bottleneck across supervised, unsupervised, and reinforcement learning.
Miscellany
- Enjoys a good puzzle
Co-authors