WISRD Benchmark: Evaluating AI Problem-Solving in Pure Image Space

HirokatuKataoka · x · 2026-08-06

This paper introduces WISRD (Image-Space Rule Discovery), a benchmark designed to evaluate the problem-solving capabilities of Image-to-Image models directly within the image space.

The research explores the potential of "Visual Intelligence" by requiring AI to read visual questions (like an IQ test), discover underlying rules, and write the answer back onto the page visually, completing an end-to-end reasoning process purely in the visual domain.

Related event: WISRD Benchmark Tests Pure Visual Reasoning in AI Models(2 posts)→

Original post →

More from Multimodal

Multimodal channel →