Sam Bowyer

I'm a Member of Technical Staff at Cohere, working on pretraining evals. Previously, I spent four years on a PhD in probabilistic machine learning at the University of Bristol's Compass CDT (viva pending!). My research interests include probabilistic modelling, deep learning, and using traditional statistics to improve LLM evaluation procedures.

Feel free to get in touch!

Sam Bowyer

Links

Papers

Efficient Benchmarking Is Just Feature Selection and Multiple Regression

2026 // preprint arxiv

Learning Generation Orders for Masked Discrete Diffusion Models via Variational Inference

2026 // ICLR DeLTa Workshop arxiv

Position: Don't Use the CLT in LLM Evals With Fewer Than a Few Hundred Datapoints

2025 // ICML Spotlight arxiv

Massively Parallel Expectation Maximization For Approximate Posteriors

2025 // AABI arxiv

Using Autodiff to Estimate Posterior Moments, Marginals and Samples

2024 // UAI arxiv

Projects

mrmr_eval

A minimal tutorial on using mRMR for efficient LLM benchmarking.

github

bayes_evals

A lightweight library for Bayesian analysis of LLM evals.

github

alan

A massively parallel probabilistic programming language.

github

Writing

Bayesian LLM Finetuning

Compass Student Blog // 2024-08-28 compass

AI UK 2025 Conference

Compass Student Blog // 2025-04-25 compass

Special Report: Compass Away Day 2024

Compass Student Blog // 2024-08-21 compass

Recommendations

Endgame

play // Samuel Beckett storygraph

La Chimera

film // Alice Rohrwacher letterboxd

Cut Worms

music // Cut Worms spotify

Elsewhere

presentations

Slides from talks and reading groups.

github

tennis

🎾🎾

Sam Bowyer // Bristol