Arip Asadulaev

A principle I first encountered as a student: you begin by making work that is bad and simple; then bad and complex; then good and complex; and finally, good and simple. This quote is often attributed to the painter Ilya Repin. I’ve found this pattern shows up far beyond painting (research, engineering, writing), and I judge ideas through this prism.
Quality versus Complexity: a learning loop Axes: Complexity increases to the right, Quality increases upward. Corners: bad & simple, bad & complex, good & complex, good & simple. Path: bad-simple to bad-complex to good-complex to good-simple. Complexity Quality bad & simple bad & complex good & complex good & simple

Working on reinforcement learning and generative models.

Good & Simple

Zero-Shot Off-Policy Learning. A. Asadulaev, M. Bobrin, S. Lahlou, D. Dylov, F. Karray, M. Takac. Preprint.

Good & Complex

Your Latent Reasoning is Secretly Policy Improvement Operator. A. Asadulaev, R. Barenjee, F. Karray, M. Takac. Preprint.

Y-Shaped Generative Flows. A. Asadulaev, S. Semyonov, A. Shtanchaev, E. Moulines, F. Karray, M. Takac. Preprint.

Bad & Complex

Rethinking Optimal Transport in Offline Reinforcement Learning. A. Asadulaev, R. Korst, A. Korotin, V. Egiazarian, A. Filchenkov, E. Burnaev. NeurIPS 2024.

Neural Optimal Transport with General Cost Functionals. A. Asadulaev, A. Korotin, V. Egiazarian, P. Mokrov, E. Burnaev. ICLR 2024.

Bad & Simple

A Minimalist Approach for Domain Adaptation with Optimal Transport. A. Asadulaev, V. Shutov, A. Korotin, A. Panfilov, V. Kontsevaya, A. Filchenkov. CoLLAs 2023.

Exploring and Exploiting Conditioning of Reinforcement Learning Agents. A. Asadulaev, I. Kuznetsov, G. Stein, A. Filchenkov. IEEE Access 2020.

See all my papers here.

Contribution beyond research:

Engineering: Designed neural computer architectures that generate molecules with favorable pharmacokinetics. These models were later validated in real-world applications.

Developed a CowSwap solver and built various AI layers to optimize token swaps, asset bridging, and other decentralized finance operations.

Education: Created and taught several courses on reinforcement learning and deep generative models, covering a range of topics from GANs to flow-matching.

Open to discussing new ideas and potential collaborations. Feel free to contact.

Social: @machinestein

Last update: Feb 6, 2026