PeopleofSuperintelligence

Researcher

Roei Herzig

Also known as Roi Herzig

Roei Herzig is a computer vision and robotics researcher who develops models and learning algorithms for Physical AI, with robots as the main testbed. He studied computer science and physics at Tel Aviv University, earned a PhD there under Amir Globerson in 2023, and won the 2023 IAAI Best PhD Thesis Award. His papers include the SKU-110K dataset in 2019, Object-Region Video Transformers in 2022, LLARVA in 2024 and ARM4R in 2025, and he was a core contributor to the Granite Vision model in 2025. As of October 2026, he is an incoming assistant professor in the School of Computing Science at Simon Fraser University, following a postdoc at UC Berkeley with Trevor Darrell.

Badges

Timeline

  1. 2015–2019

    Degrees at Tel Aviv University

    He earns a double-major BSc in computer science and physics at Tel Aviv University in 2015. A master's degree in computer science follows there in 2019, awarded magna cum laude.

  2. 2019

    The SKU-110K dataset

    In 2019, he coauthors Precise Detection in Densely Packed Scenes with Eran Goldman, Aviv Eisenschtat and colleagues, published at CVPR. The paper releases SKU-110K, an annotated dataset of packed retail environments for training and testing precise object detection.

  3. 2022

    Object-Region Video Transformers

    In 2022, he is first author of Object-Region Video Transformers (ORViT), published at CVPR with Elad Ben-Avraham, Amir Globerson, Trevor Darrell and others. ORViT extends video transformer layers with a block that feeds object representations through appearance and dynamics streams. He is also a winner of the Ego4D Point of No Return Temporal Localization Challenge at CVPR 2022.

  4. 2023

    PhD and the IAAI award

    He completes his PhD at Tel Aviv University in 2023 under Amir Globerson, with the thesis Towards Compositionality in Computer Vision. The thesis wins the 2023 IAAI Best PhD Thesis Award of the Israeli Association for Artificial Intelligence, given for the best AI thesis in Israel, and he is one of three selected for the Israeli Excellence in Data Science Postdoctoral Fellowship in 2023.

  5. 2024

    LLARVA for robot learning

    In 2024, he coauthors LLARVA with Dantong Niu, Yuvan Sharma, Trevor Darrell and others, published at CoRL. The model is pre-trained on 8.5M image-visual trace pairs from the Open X-Embodiment dataset and evaluated on 12 RLBench tasks and a physical Franka Emika Panda robot.

  6. 2025

    ARM4R and 4D representations

    In 2025, he coauthors ARM4R, published at ICML, an auto-regressive robotic model that learns low-level 4D representations from human video data. The 4D representations are 3D point tracks lifted from video with monocular depth estimation, and they transfer to robot control across several robot environments.

  7. 2025

    Core contributor to Granite Vision

    Also in 2025, he is a core contributor to Granite Vision, a lightweight, open-source multimodal model for enterprise use. Its 2-billion-parameter Granite language model is aligned with visual inputs for visual document understanding, and the model is released under the Apache-2 license.

  8. As of October 2026

    Incoming professor at SFU

    In February 2026, he gives the seminar Towards Structured Physical AI Models at UT Austin as a postdoctoral scholar at UC Berkeley, advised by Trevor Darrell, and a research scientist at the MIT-IBM Watson AI Lab. He is an incoming Assistant Professor at the School of Computing Science at Simon Fraser University. His research develops models and learning algorithms for Physical AI, using robots as the primary testbed.

Career

  1. Until 2026University of California, BerkeleyPostdoctoral Scholar
  2. Until 2026IBM ResearchResearch Scientist
  3. Simon Fraser UniversityAssistant Professor

Education

  1. 2023Tel Aviv UniversityPhD
  2. 2019Tel Aviv UniversityMaster's degree in Computer Science
  3. 2015Tel Aviv UniversityBSc in Computer Science and Physics

From the biography.

  1. More people shaping AI studied at Massachusetts Institute of Technology than at any other institutionIncludes University of California, Berkeley. We estimate where the people shaping AI studied and where their careers led. Massachusetts Institute of Technology leads with 467 former students, including 291 recorded doctorates.
  2. The company share of AI researchers’ top papers was 31% in 2020–2024 and 8% in 1970–1974Includes University of California, Berkeley. We estimate where the people shaping AI worked when their most-cited papers appeared: the company share was 31% in 2020–2024 and 8% in 1970–1974, and Google led all organizations in 2020–2024.
  3. The median h-index is 48 at universities and 14 at frontier AI labsIncludes University of California, Berkeley. The median h-index is 48 at universities and 14 at frontier AI labs. We compare publishing careers, recent citations, dual appointments and job moves across 2,420 university researchers and 308 frontier-lab researchers.
  4. Companies supply the most prior-work hires at 7 of the 9 frontier AI labsIncludes University of California, Berkeley. Our estimates of the organizations where 550 people worked directly before joining 9 frontier AI labs across 620 career transitions, comparing the main feeder organizations per lab.

Colleagues

People named in this biography, and people with dated positions at the same organization in the same years.

Sources and credits

Photo: We have no published photo of Roei Herzig yet. Send us a free-licensed photo.

  1. Roi Herzig - School of Computing Science - Simon Fraser University sfu.ca. Checked 22 October 2026.
  2. Roei Herzig - Vector Institute for Artificial Intelligence vectorinstitute.ai. Checked 22 October 2026.
  3. Precise Detection in Densely Packed Scenes arxiv.org. Checked 22 October 2026.
  4. Roei Herzig roeiherz.github.io. Checked 22 October 2026.
  5. Object-Region Video Transformers arxiv.org. Checked 22 October 2026.
  6. Prizes of IAAI iaai.org.il. Checked 22 October 2026.
  7. LLARVA: Vision-Action Instruction Tuning Enhances Robot Learning arxiv.org. Checked 22 October 2026.
  8. Pre-training Auto-regressive Robotic Models with 4D Representations arxiv.org. Checked 22 October 2026.
  9. Granite Vision: a lightweight, open-source multimodal model for enterprise Intelligence arxiv.org. Checked 22 October 2026.
  10. Towards Structured Physical AI Models | Texas ECE - Electrical & Computer Engineering at UT Austin ece.utexas.edu. Checked 22 October 2026.