REINS is proposed, a harness built on three LLM-planner couplings: LLM outputs are grounded to a learned variable-level dynamics model, verified via calibrated predicates and forward rollout, and consolidated into an indexed rule memory that serves future scenes in O(log n) without re-invoking the LLM.
Zhi-Hong Cui, Heng-Yu Liu, Hao-Ran Tang et al.· 0 citations
This work introduces Dyn-3D, a benchmark using counterfactual 3D rendering to rigorously decouple visual changes from true kinematic properties and proposes the TempoVista framework, featuring the Kinematic-GSPO algorithm.
Jiayu Ding, Zhuo-Dong Liu, Lei Zhang et al.· 0 citations
PhysVICL-74 is introduced, a framework that decomposes physically grounded VICL into physical-rule induction and transition-aligned rendering and TransPhy, a framework that decomposes physically grounded VICL into physical-rule induction and transition-aligned rendering.
Siyi Xie, Xuanke Shi, Jin-Sheng Quan et al.· 0 citations
We use cookies to run the site and, with your consent, for analytics and to show ads.
See our Cookie Policy.