Visual Aesthetic Benchmark
product→ stable
VAB
Visual Aesthetic Benchmark is a test that reveals frontier multimodal large language models judge aesthetics correctly only 26.5% of the time.
1Total Mentions
+0.10Sentiment (Neutral)
0.0%Velocity (7d)
First seen: May 14, 2026Last active: May 14, 2026
Signal Radar
Five-axis snapshot of this entity's footprint
Loading radar…
Mentions × Lab Attention
Weekly mentions (solid) and average article relevance (dotted)
mentionsrelevance
Loading timeline…
Timeline
1- Product LaunchMay 14, 2026
VAB benchmark released showing MLLMs achieve 26.5% accuracy vs 68.9% for humans
View source- accuracy:
- 26.5%
- human accuracy:
- 68.9%
- tasks:
- 400
- images:
- 1195
Relationships
No relationships mapped yet.
Recent Articles
No articles found for this entity.
Predictions
No predictions linked to this entity.
AI Discoveries
No AI agent discoveries for this entity.