Scholar
Cem Gokmen
Google Scholar ID: wCiI8oUAAAAJ
Computer Science PhD Student, Stanford University
Citations & Impact
Citations
1,110
H-index
8
i10-index
8
Publications
13
Co-authors
0
Publications
Resume (English only)
Academic Achievements
- - BEHAVIOR-1K: A human-centered embodied AI benchmark featuring 1,000 diverse, everyday activities. This four-year project addresses challenges like long-horizon tasks and complex manipulations in realistic environments.
- - BCVA: A method to predict failures in robot policies trained using Imitation Learning, enabling robots to autonomously ask for help when needed. Demonstrated 86% precision in asking for help across over 2000 trials.
- - Embodied Agent Interface: A framework for systematically benchmarking Large Language Models (LLMs) in embodied decision-making tasks. Addresses the need for standardized assessment in the rapidly evolving field of LLMs in embodied environments.
- - BEHAVIOR Vision Suite (BVS): A toolset designed to generate fully customizable synthetic data for evaluating computer vision models, addressing limitations in real-world datasets.
Research Experience
- Conducting research on Embodied AI and Robotics at the Stanford Vision & Learning Lab. Involved in multiple research projects such as BEHAVIOR-1K, BCVA (Behavioral Cloning Value Approximation), and Embodied Agent Interface.
Education
- Pursuing a Ph.D. in Computer Science at Stanford University, advised by Prof. Fei-Fei Li.
Background
- Currently a CS PhD Candidate at Stanford, specializing in Embodied AI and Robotics. Research interests include developing machine learning algorithms to enable robots to perform complex, long-horizon tasks in real-world environments using techniques like learning from demonstrations, reinforcement learning, sim-to-real transfer, and leveraging off-the-shelf LLM/VLM models. Also has extensive experience building simulation environments and benchmarks to support this research.
Co-authors