P. A. Zhokhov

Papers

Total Citations

H-Index

About

P. A. Zhokhov is a leading researcher in artificial intelligence, specializing in sequential decision-making, imitation learning, and large-scale pretraining for embodied agents. Their most notable contribution is the development of Video PreTraining (VPT), a groundbreaking framework that enables AI agents to learn complex behaviors by watching unlabeled online videos. This work, published in 2022 and garnering 50 citations, demonstrates how internet-scale, noisy video data can be leveraged to pretrain models for robotics, video games, and computer use—domains where labeled data is scarce. By bridging the gap between passive video observation and active task execution, Zhokhov has opened new pathways for training generalist agents without expensive human demonstrations. Their research sits at the intersection of computer vision, reinforcement learning, and large-scale machine learning, offering a scalable solution to one of AI’s hardest challenges: learning from unstructured, real-world data. Zhokhov’s work has significant implications for autonomous systems, making them more adaptable and data-efficient, and positions them as a key innovator in the quest for broadly capable AI.