Samuel J. Bell

AI Research Scientist

profile-august.jpg

I’m Sam 👋, a Research Scientist at FAIR, the fundamental AI research group of Meta’s Superintelligence Labs. I split my time between London and Cambridge, UK.

My research is broadly focused on both evaluating and improving the reliability of LLMs. I’m particularly interested in understanding model performance in novel or out-of-distribution scenarios, such as prompts unlike those seen during pre- or post-training. I’m also very excited by techniques for better handling uncertainty, ranging from missing training data to ambiguous user intent.

I joined FAIR London in 2023 after a postdoc with FAIR Paris. In 2022 I completed my PhD in machine learning as part of the ML@CL group at the University of Cambridge, supervised by Prof. Neil Lawrence. Previously, I have studied at The Alan Turing Institute, obtained a master’s in natural language processing at the Cambridge Computer Laboratory, and did my bachelor’s in computer science at the University of Manchester.

In between, I’ve simulated financial crises in market risk at Goldman Sachs, built new retail banks at Thought Machine, and developed next generation credit scores at Credit Kudos.

If you’re interested in learning more about what I do, or discussing research collaborations or opportunities, please do drop me a line. My email is samueljamesbell [at] gmail [dot] com. I’d love to hear from you.

Selected publications

  1. abstention-bench.png
    AbstentionBench: Reasoning LLMs Fail on Unanswerable Questions
    Polina Kirichenko*, Mark Ibrahim*, Kamalika Chaudhuri, and Samuel J. Bell*
    In review. *Equal contribution, 2025
  2. toxicity.png
    On the Role of Speech Data in Reducing Toxicity Detection Bias
    Samuel J. Bell, Megan Richards, Eduardo Sanchez, Mariano Coria Meglioli, and 4 more authors
    In North American Chapter of the Association for Computational Linguistics (NAACL), 2025
  3. spurious-correlations-benchmarks.png
    Reassessing the Validity of Spurious Correlations Benchmarks
    Samuel J. Bell, Diane Bouchacourt, and Levent Sagun
    2024
  4. sketch-distribution-full-with-group.png
    Simplicity Bias Leads to Amplified Performance Disparities
    Samuel J. Bell, and Levent Sagun
    In ACM Conference on Fairness, Accountability, and Transparency (FAccT), 2023
  5. svm-final-test-acc-mean-ivr.png
    Modeling the Machine Learning Multiverse
    Samuel J. Bell, Onno Kampman, Jesse Dodge, and Neil Lawrence
    In Advances in Neural Information Processing Systems (NeurIPS), 2022