# ThinkWorld: World Imagination Reinforces Chain-of-Dynamics Reasoning > Source: > Published: 2026-10-28 02:51:35+00:00 Vision-Language-Action (VLA) models have made substantial progress in robotic manipulation, but their generalization remains constrained by the diversity and coverage of action-labeled robot demonstrations. Human videos provide much broader behavioral and environmental experience, while learning fro