RUS  ENG
Full version
JOURNALS // Artificial Intelligence and Decision Making // Archive

Artificial Intelligence and Decision Making, 2026 Issue 2, Pages 93–106 (Mi iipr673)

Machine learning, neural networks

Finetuning of object-centric models for reinforcement learning

D. V. Vasileva, A. I. Panovab

a Moscow Institute of Physics and Technology (National Research University), Dolgoprudny, Russia
b Federal Research Center "Computer Science and Control" of Russian Academy of Sciences, Moscow

Abstract: Object-centric models represent an image as a set of vectors describing the objects present in the image. This image representation allows for explicit modeling of interactions between objects, which accelerates the convergence of reinforcement learning algorithms. A drawback of using such models in reinforcement learning tasks is their typical use in a frozen state, which can lead to problems caused by distributional shift from the data on which these models were originally trained. This paper presents an algorithm for fine-tuning object-centric models by splitting the training procedure into a model finetuning phase and a policy learning phase. Experiments confirm the effectiveness of this approach, although there exist scenarios where this method underperforms compared to using frozen models.

Keywords: object-centric models, reinforcement learning, self-supervised learning, finetuning.

DOI: 10.14357/20718594260208



Bibliographic databases:


© Steklov Math. Inst. of RAS, 2026