Uliana Vinogradova

Russian State Scientific Center for Robotics and Technical Cybernetics

Papers

Total Citations

H-Index

About

Uliana Vinogradova is a rising researcher at the intersection of robotics, natural language processing, and computer vision. Her work centers on enhancing the adaptability and robustness of autonomous systems, particularly through the integration of large language models (LLMs) and visual feedback. Her most notable contribution, the paper "LERa: Replanning with Visual Feedback in Instruction Following" (2025), introduces a novel framework that enables robots to dynamically adjust their plans in response to real-world changes. By combining visual observations with language-based reasoning, LERa—short for Look, Explain, Replan—allows agents to detect failures, explain discrepancies, and generate corrective actions without human intervention. This work addresses a critical limitation in LLM-driven robotics: the inability to handle unexpected environmental shifts using only textual inputs. Though early in her career, Vinogradova’s research has already garnered attention for its practical implications in autonomous task execution. Her approach promises to make robotic systems more resilient and context-aware, paving the way for safer, more flexible deployment in dynamic settings. With a clear focus on bridging perception and language, Vinogradova is a promising voice in the future of intelligent robotics.