ainewsblitz.com

Breaking

NVIDIA Research Unveils Spatial-IQ, a Diagnostic Benchmark Exposing How Badly Multimodal Models Count 3D Objects

NVIDIA Research Unveils Spatial-IQ, a Diagnostic Benchmark Exposing How Badly Multimodal Models Count 3D Objects
  • Research & Papers
  • Foundation Models
  • Robotics

NVIDIA Research has released Spatial-IQ, a diagnostic benchmark that decomposes 3D object counting—including hidden objects—into nine perceptual and cognitive sub-tasks, revealing a stark gap between humans and today's best multimodal systems. On the headline task of counting objects in a stacked 3D structure, humans reach 82.1% accuracy, while the strongest off-the-shelf multimodal large language model manages just 17.7%.

Continue reading

The rest of this article is for AI News Blitz readers. Choose an option below to keep reading.

$20
Read this article
$29/month
Unlimited — all 27,934 articles, the full archive, and comprehension quizzes
Save 72%
$98/year
≈ $8.17/month
Unlimited, billed once a year