Dewi S.W. Gould

Researcher, Arcadia Impact Alignment

prof_pic.jpg

I am a research lead on the Arcadia Impact Alignment Team. I’m thinking about scalable oversight and automated alignment research.

Previously I…

  • Was an Astra Fellow at Redwood Research, working on general purpose elicitation and no-CoT time horizons.
  • Was a post-doc at the Alan Turing Institute on Project Bluebird, working on synthetic scenario generation for autonomous air-traffic control. I gave a Pint of Science talk about it.
  • Did a PhD in mathematical physics at the University of Oxford. My thesis was on Generalized Symmetries in String Theory Realizations of Quantum Field Theories. I wrote a blog post about my research.

news

Oct 27, 2026 New blog post AI speeds up 3rd party safety researchers by ~1.6x (via self-reports) on LessWrong.
Oct 06, 2026 New blog post Benchmarking Jev against no-CoT LLMs on LessWrong.
Oct 01, 2026 New blog post Lessons from Building an Automated Research Scaffold on LessWrong.

recent research

  1. speedup.png
    AI speeds up 3rd party safety researchers by ~1.6x (via self-reports)
    Angel Martinez, Alejandro Aristizabal, Nicole Xing, and 1 more author
    2026
    Blog post, LessWrong
  2. jev.png
    Benchmarking Jev against no-CoT LLMs
    Dewi S. W. Gould, and James Mann
    2026
    Blog post, LessWrong
  3. scaffold.png
    Lessons from Building an Automated Research Scaffold
    Alejandro Aristizabal, Josh Hills, Dewi S. W. Gould, and 4 more authors
    2026
    Blog post, LessWrong
  4. oversight.png
    Obstacles to the Scalable Oversight of Auto-Alignment Research
    Sam Martin, Dewi S. W. Gould, Cameron Holmes, and 1 more author
    2026
    Blog post, LessWrong
  5. astra.png
    Estimating GPT-6 Astra’s No-CoT Time Horizon
    Francis Rhys Ward, and Dewi S. W. Gould
    2026
    Blog post, LessWrong
  6. debate.png
    Debate with Self-Play Best-of-N Optimization
    Dewi S. W. Gould, Sam Martin, Alejandro Aristizabal, and 2 more authors
    2026
    Blog post, LessWrong
  7. think_fast.png
    Think Fast: Estimating No-CoT Task-Completion Time Horizons of Frontier AI Models
    Dewi S. W. Gould, Francis Rhys Ward, Anders Cairns Woodruff, and 18 more authors
    arXiv preprint, 2026
  8. positive_case.png
    A Positive Case for Faithfulness: LLM Self-Explanations Help Predict Model Behavior
    H. Mayne, J. S. Kang, Dewi S. W. Gould, and 3 more authors
    In Proceedings of the 43rd International Conference on Machine Learning (ICML), 2026
  9. skate.png
    SKATE, a Scalable Tournament Eval: Weaker LLMs differentiate between stronger ones using verifiable challenges
    Dewi S. W. Gould, B. Mlodozeniec, and S. F. Brown
    arXiv preprint, 2025
  10. pac.png
    PAC Apprenticeship Learning with Bayesian Active Inverse Reinforcement Learning
    O. Bajgar, Dewi S. W. Gould, J. Liu, and 3 more authors
    In Reinforcement Learning Conference, 2025
    Earlier version at NeurIPS 2024 Workshop on Bayesian Decision-making and Uncertainty