Wikibench

MuirBench

reasoning

A comprehensive benchmark for robust multi-image understanding capabilities of multimodal LLMs. Consists of 12 diverse multi-image tasks involving 10 categories of multi-image relations (e.g., multiview, temporal relations, narrative, complementary). Comprises 11,264 images and 2,600 multiple-choice questions created in a pairwise manner, where each standard instance is paired with an unanswerable variant for reliable assessment.

Leaderboard

Showing 11 of 11 results

  1. Qwen3 VL 32B Thinking
    80.3%
    i
  2. Qwen3 VL 235B A22B Thinking
    80.1%
    i
  3. Qwen3 VL 30B A3B Thinking
    77.6%
    i
  4. Qwen3 VL 8B Thinking
    76.8%
    i
  5. Qwen3 VL 4B Thinking
    75.0%
    i
  6. Qwen3 VL 235B A22B Instruct
    72.8%
    i
  7. Qwen3 VL 32B Instruct
    72.8%
    i
  8. Qwen3 VL 8B Instruct
    64.4%
    i
  9. Qwen3 VL 4B Instruct
    63.8%
    i
  10. Qwen3 VL 30B A3B Instruct
    62.9%
    i
  11. Qwen2.5-Omni-7B
    59.2%
    i