Tatiana Zemskova

Moscow Institute of Physics and Technology

Papers

Total Citations

H-Index

About

Tatiana Zemskova is an emerging researcher specializing in 3D scene understanding, open-vocabulary object grounding, and autonomous agent perception — fields that sit at the intersection of computer vision, natural language processing, and robotics. Her most notable work, "Beyond Bare Queries: Open-Vocabulary Object Grounding with 3D Scene Graph," addresses a critical limitation in existing CLIP-based methods: their inability to handle ambiguous natural language descriptions that require contextual reasoning about object relationships within a scene. By leveraging 3D scene graphs, Zemskova's approach enables autonomous agents to locate objects described through complex, relational queries rather than simple, direct commands — a meaningful step toward more human-like spatial reasoning in AI systems. The paper has garnered 13 citations across its 2024 and 2025 versions, reflecting strong and rapid uptake within the research community since its publication. For students and researchers working on embodied AI, robotics navigation, or language-grounded perception, Zemskova's contributions offer an important foundation for understanding how structured scene representations can bridge the gap between natural language ambiguity and precise 3D object localization.