Troshin, Sergey

4 publications

TMLR 2025 On the Low-Rank Parametrization of Reward Models for Controlled Language Generation Sergey Troshin, Vlad Niculae, Antske Fokkens

ICLR 2023 CodeBPE: Investigating Subtokenization Options for Large Language Model Pretraining on Source Code Nadezhda Chirkova, Sergey Troshin

TMLR 2023 Wrapped $\beta$-Gaussians with Compact Support for Exact Probabilistic Modeling on Manifolds Sergey Troshin, Vlad Niculae

ICLRW 2022 CodeBPE: Investigating Subtokenization Options for Large Language Model Pretraining on Source Code Nadezhda Chirkova, Sergey Troshin