Sasha Voitovych

I am a second-year PhD student at MIT IDSS / LIDS, very fortunate to be advised by Alexander (Sasha) Rakhlin. My interests are in the foundations of interactive learning. Recently, I have been working on oracle-efficient online learning and reinforcement learning for language model post-training.

During my undergrad at the University of Toronto, I was very lucky to work with Daniel Roy, Rachid Guerraoui, and Dan Alistarh.

Links: CV, Google Scholar, LinkedIn.

Highlights

I was named an Amazon AI Fellow 🎓 for the 2025–2026 and 2026–2027 academic years.
Our paper on traceability in SCO received a Spotlight ⭐ at NeurIPS 2025.

Publications

Learning with Simulators: No Regret in a Computationally Bounded World
Sasha Voitovych*, Abhishek Shetty*, Noah Golowich, Alexander Rakhlin
Conference on Learning Theory (COLT), 2026
[arXiv]
On Traceability in $\ell_p$ Stochastic Convex Optimization
Sasha Voitovych*, Mahdi Haghifam*, Idan Attias, Gintare K. Dziugaite, Roi Livni, Daniel Roy
Conference on Neural Information Processing Systems (NeurIPS), 2025
⭐ Spotlight Presentation
[arXiv]
Byzantine-Robust Federated Learning: Impact of Client Subsampling and Local Updates
Youssef Allouah, Sadegh Farhadkhani, Rachid Guerraoui, Nirupam Gupta, Rafael Pinot, Geovani Rizk, Sasha Voitovych
International Conference on Machine Learning (ICML), 2024
[arXiv]
On the Inherent Anonymity of Gossiping
Rachid Guerraoui, Anne-Marie Kermarrec, Anastasiia Kucherenko, Rafael Pinot, Sasha Voitovych
International Symposium on Distributed Computing (DISC), 2023
[arXiv]
Near-Optimal Leader Election in Population Protocols on Graphs
Dan Alistarh, Joel Rybicki, Sasha Voitovych
ACM Symposium on Principles of Distributed Computing (PODC), 2022
[arXiv]