Scott Emmons I am a research scientist at Google DeepMind focused on AI safety and alignment. I am wrapping up my PhD at UC Berkeley’s Center for Human-Compatible AI, advised by Stuart Russell. I previously cofounded far.ai, a 501(c)3 research nonprofit that incubates and accelerates beneficial AI research agendas. I am interested in both the theory and practice of AI alignment. I have helped characterize how RLHF can lead to deception when the AI sees more than the human, develop multimodal attacks and benchmarks for open-ended agents, and use mechanistic interpretability to find evidence of learned look-ahead in a chess-playing neural network. Curriculum Vitae scott at scottemmons dot com

Publications

• When Do Universal Image Jailbreaks Transfer Between Vision-Language Models?
  Rylan Schaeffer, Dan Valentine, Luke Bailey, James Chua, Cristóbal Eyzaguirre, Zane Durante, Joe Benton, Brando Miranda, Henry Sleight, John Hughes, Rajashree Agrawal, Mrinank Sharma, Scott Emmons, Sanmi Koyejo, and Ethan Perez
  arXiv, 2024
  [arXiv] [BibTeX]
• Evidence of Learned Look-Ahead in a Chess-Playing Neural Network
  Erik Jenner, Shreyas Kapur, Vasil Georgiev, Cameron Allen, Scott Emmons, and Stuart Russell
  arXiv, 2024
  [project page] [code] [arXiv] [BibTeX]
• When Your AIs Deceive You: Challenges of Partial Observability in Reinforcement Learning from Human Feedback
  Leon Lang, Davis Foote, Stuart Russell, Anca Dragan, Erik Jenner, and Scott Emmons
  arXiv, 2024
  [arXiv] [BibTeX]
• A StrongREJECT for Empty Jailbreaks
  Alexandra Souly^*, Qingyuan Lu^*, Dillon Bowen^*, Tu Trinh^†, Elvis Hsieh^†, Sana Pandey, Pieter Abbeel, Justin Svegliato, Scott Emmons, Olivia Watkins, and Sam Toyer
  arXiv, 2024
  [code] [arXiv] [BibTeX]
• ALMANACS: A Simulatability Benchmark for Language Model Explainability
  Edmund Mills, Shiye Su, Stuart Russell, and Scott Emmons
  arXiv, 2023
  [code] [arXiv] [BibTeX]
• Image Hijacks: Adversarial Images can Control Generative Models at Runtime
  Luke Bailey^*, Euan Ong^*, Stuart Russell, and Scott Emmons
  International Conference on Machine Learning, 2024
  [project page] [code] [arXiv] [BibTeX]
• Do the Rewards Justify the Means? Measuring Trade-Offs Between Rewards and Ethical Behavior in the MACHIAVELLI Benchmark
  Alexander Pan^*, Chan Jun Shern^*, Andy Zou^*, Nathaniel Li, Steven Basart, Thomas Woodside, Jonathan Ng, Hanlin Zhang, Scott Emmons, and Dan Hendrycks
  International Conference on Machine Learning, 2023
  [project page] [code] [arXiv] [BibTeX]
• For Learning in Symmetric Teams, Local Optima are Global Nash Equilibria
  Scott Emmons, Caspar Oesterheld, Andrew Critch, Vince Conitzer, and Stuart Russell
  International Conference on Machine Learning, 2022
  [code] [arXiv] [BibTeX]
• RvS: What is Essential for Offline RL via Supervised Learning?
  Scott Emmons, Benjamin Eysenbach, Ilya Kostrikov, and Sergey Levine
  International Conference on Learning Representations, 2022
  [code] [arXiv] [BibTeX]
• An Empirical Investigation of Representation Learning for Imitation
  Xin Chen^*, Sam Toyer^*, Cody Wild^*, Scott Emmons, Ian Fischer, Kuang-Huei Lee, Neel Alex, Steven H. Wang, Ping Luo, Stuart Russell, Pieter Abbeel, and Rohin Shah
  Neural Information Processing Systems, 2021
  [code] [arXiv] [BibTeX]
• Sparse Graphical Memory for Robust Planning
  Scott Emmons^*, Ajay Jain^*, Michael Laskin^*, Thanard Kurutach, Pieter Abbeel, and Deepak Pathak
  Neural Information Processing Systems, 2020
  [video] [code] [arXiv] [BibTeX]
• Concurrency and Reachability in Treelike Temporal Networks
  Eun Lee, Scott Emmons, Ryan Gibson, James Moody, and Peter J. Mucha
  Physical Review E, 2019
  [arXiv] [BibTeX]
• A Map Equation with Metadata: Varying the Role of Attributes in Community Detection
  Scott Emmons and Peter J. Mucha
  Physical Review E, 2019
  [code] [arXiv] [BibTeX]
• Global Redundancy Resolution via Continuous Pseudoinversion of the Forward Kinematic Map
  Kris Hauser and Scott Emmons
  IEEE Transactions on Automation Science and Engineering, 2018
  [project page] [code] [preprint] [BibTeX]
• MOOC Visual Analytics: Empowering Students, Teachers, Researchers, and Platform Developers of Massively Open Online Courses
  Scott Emmons, Robert Light, and Katy Börner
  Journal of the Association for Information Science and Technology (JASIST), 2017
  [project page] [code] [preprint] [BibTeX]
• Post-Processing Partitions to Identify Domains of Modularity Optimization
  William H. Weir, Scott Emmons, Ryan Gibson, Dane Taylor, and Peter J. Mucha
  Algorithms, 2017
  [code] [arXiv] [BibTeX]
• Analysis of Network Clustering Algorithms and Cluster Quality Metrics at Scale
  Scott Emmons, Stephen Kobourov, Mike Gallant, and Katy Börner
  PLoS ONE, 2016
  [project page] [code] [arXiv] [BibTeX]

Open-Source Software

• imitation: Clean Imitation Learning Implementations
  Adam Gleave, Mohammad Taufeeque, Juan Rocamonde, Erik Jenner, Steven H. Wang, Sam Toyer, Maximilian Ernestus, Nora Belrose, Scott Emmons, and Stuart Russell
  arXiv, 2022
  [code] [arXiv] [documentation] [BibTeX]

Leadership

• Sysadmin and Internship Manager, Center for Human-Compatible AI.
  Berkeley, CA, August 2019 to May 2024.
  [about]
  
• Cofounder and President, far.ai.
  Berkeley, CA, February 2022 to July 2023.
  [about]
  

Service

• Volunteer Teacher, Shanti Bhavan Children's Project.
  Tamil Nadu, India, Summer 2017.
  [about] [Netflix documentary]
  
• Volunteer Teacher, Sunflower County Freedom Project.
  Mississippi Delta, Summer 2016.
  [about]