I am a computer science researcher at The University of North Carolina at Chapel Hill. My research focuses on embodied agentic systems and computer vision, with a particular emphasis on building intelligent agents that can perceive, reason about, and interact with the physical world. I am interested in how large vision-language models can be leveraged to enable embodied AI agents to perform complex tasks in dynamic environments.
🔥 News
- 2026.03: 🎉🎉 Prune-Then-Plan accepted to CVPR 2026 Findings!
- 2024.05: 🎉🎉 Accepted to the PhD program in Computer Science at UNC Chapel Hill!
- 2024.05: 🎉🎉 Completed Masters of Science in Computer Science at UNC
- 2025.05: 🎉🎉 Released VIN-NBV on arXiv
- 2024.04: 🎉🎉 Monitor Illumination paper accepted to the CVPR 2024 Workshop on Multimedia Forensics!
- 2023.01: 🎉🎉 Started Masters of Science in Computer Science at UNC!
- 2023.12: 🎉🎉 Completed Bachelor of Science in Computer Science at UNC
📝 Publications
Noah Frahm, Prakrut Patel, Yue Zhang, Shoubin Yu, Mohit Bansal, Roni Sengupta
- We propose Prune-Then-Plan, a framework that stabilizes VLM-driven embodied exploration through step-level calibration. Our method prunes implausible frontier choices using a Holm-Bonferroni inspired pruning procedure and delegates final decisions to a coverage-based planner, achieving up to 49% and 33% relative improvements in visually grounded SPL and LLM-Match metrics.
VIN-NBV: A View Introspection Network for Next-Best-View Selection
Noah Frahm, Dongxu Zhao, Andrea Dunn Beltran, Ron Alterovitz, Jan-Michael Frahm, Junier Oliva, Roni Sengupta
- We introduce the View Introspection Network (VIN), a lightweight neural network that predicts the Relative Reconstruction Improvement of a potential next viewpoint without making new acquisitions. VIN-NBV achieves ~30% gain in reconstruction quality over coverage-based criteria and outperforms deep RL methods by ~40%.
Building Secure and Engaging Video Communication by Using Monitor Illumination
Jun Myeong Choi, Johnathan Chi-Ho Leung, Noah Frahm, Max Christman, Gedas Bertasius, Roni Sengupta
- We use light reflected from the monitor to detect if a person in a video call is real/live (on) or deepfake (off).
📖 Education
- 2020.08 - 2023.12, Bachelor of Science Computer science, The University of North Carolina.
- 2023.01 - 2024.05, Masters of Science Computer Science, The University of North Carolina
- 2024.05 - (now), Doctor of philosphy Computer Science, The University of North Carolina
💻 Internships
- Summer 2024, Applied Scientist, EveryPoint
- Summer 2023, Software Engineering Intern, Capital One
- Summer 2022, Software Engineering Intern, Capital One