SpatialBrain is a framework that retains reasoning‑decision experience while keeping its vision‑language backbone fixed. It separates raw scene observation from judgments about when reconstruction, measurement, or reference‑frame changes are needed.
3DFusionMemory fuses current‑scene observations into queryable spatial relations. A Spatial Skill Library offers reusable tool procedures, and Spatial Cognition links task conditions with experience on how and when to employ memory and skills.
Completed trajectories are abstracted into process records and consolidated across task blocks, allowing later decisions to draw on earlier execution without transferring scene coordinates or reference answers.
Across five spatial‑reasoning benchmarks, SpatialBrain outperforms direct answers from the same backbone on four benchmarks and ties on the fifth. Component comparisons, experience‑coverage curves, and traced resource decisions illustrate how the framework accumulates and utilizes these resources.
Review: By explicitly modeling experience, the framework enables flexible reuse of memory and tools in complex spatial tasks, improving reasoning consistency and efficiency across tasks.