Publications (8)
Transfer Learning with Deep Tabular Models
Roman Levin, Valeriia Cherepanova, Avi Schwarzschild +5
Recent work on deep learning for tabular data demonstrates the strong performance of deep tabular models, often bridging the gap between gradient boosted decision trees and neural…
Echo Chambers in Collaborative Filtering Based Recommendation Systems
Emil Noordeh, Roman Levin, Ruochen Jiang +1
Recommendation systems underpin the serving of nearly all online content in the modern age. From Youtube and Netflix recommendations, to Facebook feeds and Google searches, these s…
On the Riemann-Hilbert Problem for Difference and -Difference Systems
Ilya Vyugin, Roman Levin
In this paper we study an analogue of the classical Riemann-Hilbert problem stated for the classes of difference and -difference systems. The Birkhoff's existence theorem was ge…
Has My System Prompt Been Used? Large Language Model Prompt Membership Inference
Roman Levin, Valeriia Cherepanova, Abhimanyu Hans +2
Prompt engineering has emerged as a powerful technique for optimizing large language models (LLMs) for specific applications, enabling faster prototyping and improved performance,…
Study of plasma heating in ohmically and auxiliary heated regimes in spherical tokamak Globus-M
Nikolay Sakharov, Bayr Ayushin, Alexander Barsukov +11
The ion temperature behavior in the plasma core of the spherical tokamak Globus-M (major radius 0.36 m, minor radius 0.24 m, torus aspect ratio 1.5, toroidal magnetic field near th…
Where do Models go Wrong? Parameter-Space Saliency Maps for Explainability
Roman Levin, Manli Shu, Eitan Borgnia +3
Conventional saliency maps highlight input features to which neural network predictions are highly sensitive. We take a different approach to saliency, in which we identify and ana…