AI TrackerFrontier desk · AEST
Back to news
normalaitech

MindView: Benchmarking Generative Multi-View Spatial Intelligence

arXiv cs.AI · 21 Oct 2026, 03:58 AEST

MindView is a new benchmark for generative multi-view spatial intelligence, featuring 1,500 samples from 513 scenes. It evaluates eleven generators across seven tasks, revealing significant opportunities for improvement in spatially grounded view generation.

Key facts

Read aloud

MindView is a new benchmark for generative multi-view spatial intelligence, featuring one thousand five hundred samples from five hundred thirteen scenes. It evaluates eleven generators across seven tasks, revealing significant opportunities for improvement in spatially grounded view generation. Spatial intelligence has recently been assessed through generative tasks such as spatially grounded image editing, yet language-guided generation of spatially consistent views from partial multi-view observations remains underexplored. We introduce MindView, a benchmark for generative multi-view spatial intelligence with one thousand five hundred samples from five hundred thirteen scenes across three indoor datasets. Evaluation of eleven generators reveals substantial room for improvement on MindView. A paired context analysis further shows that all eleven generators obtain lower overall scores under the expanded context setting. Overall, MindView provides a concrete testbed for generative multi-view spatial intelligence.

Sources in this cluster