NVIDIA and Yale Introduce Spatial-IQ Benchmark

NVIDIA and Yale introduced the Spatial-IQ benchmark to evaluate the 3D spatial reasoning capabilities of MLLMs. Tests reveal that top models achieve only 17.7% accuracy in 3D object counting, significantly lagging behind humans at 82.1%.

2026-08-01 ~ 2026-08-01 · 2 related posts

• NVIDIA's Spatial-IQ Benchmark Exposes Multimodal Models' Flaws in 3D Reasoning — NVIDIAAI · 2026-08-01
• NVIDIA's Spatial-IQ Benchmark Exposes MLLM 3D Counting Flaws (17.7% vs 82.1% Human Accuracy) — NVIDIAAI · 2026-08-01