ZeroBench: An Impossible Visual Benchmark for Contemporary Large Multimodal Models

TL;DR

ZeroBench is a challenging visual reasoning benchmark designed to be impossible for current large multimodal models, highlighting their limitations and encouraging future improvements in visual understanding.

Contribution

We introduce ZeroBench, a novel lightweight benchmark with impossible questions for current LMMs, and provide a comprehensive evaluation and analysis of their failures.

Findings

All evaluated LMMs scored 0.0% on ZeroBench.

ZeroBench remains impossible despite ongoing model progress.

The benchmark is publicly available to foster future research.

Abstract

Large Multimodal Models (LMMs) exhibit major shortfalls when interpreting images and, by some measures, have poorer spatial cognition than small children or animals. Despite this, they attain high scores on many popular visual benchmarks, with headroom rapidly eroded by an ongoing surge of model progress. To address this, there is a pressing need for difficult benchmarks that remain relevant for longer. We take this idea to its limit by introducing ZeroBench-a lightweight visual reasoning benchmark that is entirely impossible for contemporary frontier LMMs. Our benchmark consists of 100 manually curated questions and 334 less difficult subquestions. We evaluate 20 LMMs on ZeroBench, all of which score 0.0%, and rigorously analyse the errors. To encourage progress in visual understanding, we publicly release ZeroBench.