WROP: Open-Source Suite Trains Object Permanence in World Models

Researchers released WROP, an open-source data infrastructure with 150 cognitive tasks and about 1.5 million samples—including over 10,000 generated with Blender—to systematically train object permanence in world models (arXiv:2609.28654).

2026-09-25 ~ 2026-09-26 · 3 related posts

• WROP: 150 cognitive tasks train object permanence into video world models — Haotian Zhang · 2026-09-25
• Blender-generated datasets teach world models object permanence at 10k samples per task — _akhaliq · 2026-09-26
• Open data infrastructure teaches AI object permanence: 150 tasks, 1.5M samples, 16B model — du_yilun · 2026-09-26