Scholar
Koustuv Sinha
Google Scholar ID: 9P9QcckAAAAJ
Research Scientist, Meta AI (Fundamental AI Research), McGill University (MSc, PhD)
Citations & Impact
Citations
3,000
H-index
21
i10-index
31
Publications
20
Co-authors
35
list available
Publications
Resume (English only)
Academic Achievements
- Contributed to V-JEPA 2, a frontier self-supervised video model enabling understanding, prediction, and planning.
- Co-authored 'Scaling Language-Free Visual Representation Learning'.
- Contributed to MetaMorph, a multimodal understanding and generation model via instruction tuning.
- Worked on VEDIT, a latent prediction architecture for procedural video representation learning.
- Part of the Chameleon team, developing mixed-modal early-fusion foundation models; code and weights publicly released.
- Research covered by Nature, VentureBeat, InfoQ, DailyMail, and Hindustan Times.
- Served as Senior Area Chair at ACL 2024.
- Released the Physical World Reasoning leaderboard and MVPBench dataset for evaluating video understanding in VLMs.
