MuirBench: leaderboard

2,600 multiple-choice questions over 11,264 images across 12 multi-image tasks and 10 image-relation types, from USC, Microsoft Research and partners (2024); each item has an unanswerable twin.

Source: arxiv.org.

Interactive version: theaggregate.ai/benchmark?slug=muirbench · How It Works · Data refreshed daily, snapshot 2026-09-05.