Vstar Bench

'Vstar Bench' is a benchmark used to evaluate the performance of models on vision-centric problems, specifically assessing their ability to handle complex visual reasoning tasks synthesized from a diverse dataset of over 1 million high-quality problems.

Papers using Vstar Bench (3)

Q-probe: Scaling Image Quality Assessment To High Resolution Via Context-aware Agentic Probing2026VISTA-Bench: Do Vision-Language Models Really Understand Visualized Text as Well as Pure Text?2026Long Grounded Thoughts: Synthesizing Visual Problems and Reasoning Chains at Scale2025