Martin Marek

I am a Member of Technical Staff at Recursive.
You can find some examples of my work below.

Links

GitHub, Google Scholar, Twitter

Email me: martin.m@nyu.edu

Papers

Score Centering Stabilizes
Off-policy Reinforcement Learning
Martin Marek & Max Ryabinin
[arXiv, PDF, GitHub, Twitter]
Forgetting in Language Models:
Capacity, Optimization, and Self-Generated Replay
M. Marek, D. Cho, S. Qiu, R. Chunara, P. Izmailov, A. G. Wilson
[arXiv, PDF, GitHub, Twitter]
Small Batch Size Training for Language Models:
When Vanilla SGD Works, and Why Gradient Accumulation Is Wasteful
M. Marek, S. Lotfi, A. Somasundaram, A. G. Wilson, M. Goldblum
[arXiv, PDF, GitHub, Twitter, Blog]
Can a Confident Prior Replace a Cold Posterior?
Martin Marek, Brooks Paige, Pavel Izmailov
[arXiv, PDF, GitHub, Twitter]

Personal projects

Burst Photo
Mac app that brings “night mode” to any camera.
Implements HDR+ from scratch in Metal Shading Language.
[Website, GitHub, DPReview, PetaPixel, SonyAlphaRumors]

Website design inspired by Pavel Izmailov.