Hello! I'm a researcher focused on continual adaptation, generalization and efficient exploration & learning. I got BSc & MSc in Mathematics at the Higher School of Economics, Moscow. Right now doing PhD at Computational Imaging Lab.
This space is where I share my latest research, publications, and my thoughts regarding new ideas.
Preprint
The paper addresses the problem of zero-shot adaptation of Behavioral Foundational Models (BFMs), where a policy extracted at test time is far from optimal. We propose a training-free method which finds a close to optimal policy at test time by leveraging distribution correction coefficient extracted from the pretrained BFM.
ICLR 2026 (Poster)
We present a method for zero-shot adapation of BFMsto unseen environmental dynamics. Previously, those models were not able to adapt to unseen dynamics. We propose to learn a belief state transformer to adapt BFMs to handle even completely OOD cases.
ICLR 2026 (Poster)
We present Hamiltonian Optimal Transport Advection (HOTA), a Hamilton-Jacobi-Bellman based method that tackles the dual dynamical OT problem explicitly through Kantorovich potentials, enabling efficient and scalable trajectory optimization.Empirically, HOTA outperforms all baselines in standard benchmarks, as well as in custom datasets with non-differentiable costs, both in terms of feasibility and optimality.
NeurIPS 2024 (Spotlight)
We present a new approach for Neural Optimal Transport (NOT) training procedure, capable of accurately and efficiently estimating optimal transportation plan via specific expectile regularization on dual Kantorovich potentials. Proposed method, called Expectile-Regularized Neural Optimal Transport (ENOT), outperforms previous state-of-the-art approaches in the established benchmarks.