Holi-World turns egocentric video into 3D/4D data: 4B model beats zero-shot 27B
liuziwei7 · x · 2026-10-24
- Holi-World by Ropedia converts everyday egocentric video into 3D and 4D world data with no manual labels.
- From 100 self-collected scenes it yields static 3D objects, dynamic 4D tracks, and 80K grounded QA pairs.
- A 4B model trained on this data outperforms a zero-shot 27B on spatial and temporal reasoning, pointing to a scalable self-supervised world-data pipeline.
More from Embodied
- microagi promises a billion robots by 2029; skeptics say scene understanding won't be solved by 2027 — AlexTensor · 2026-10-24
- Tesla registers 69 more Cybercabs, bringing its Robotaxi fleet to 420 vehicles — mitchdeg · 2026-10-24
- CMU RI PhD open-sources full robotics faculty job market application materials — sh_reya · 2026-10-24
- Ropedia A0: code-only robot policy hits 70.24% on RoboDojo, 75% first-try sim2real — liuziwei7 · 2026-10-24
- Unitree's Wang Xingxing: GPT-6 Astra handles robot painting and Go with no specialized training — Afinetheorem · 2026-10-24
- Rodney Brooks on Amazon's 1M robots: real-world deployment is the hard part — rodneyabrooks · 2026-10-24