MIABench

multimodal

MIABench evaluates multimodal instruction alignment and following capabilities.

Methodology

Imported from llm-stats public benchmark metadata. Modality: multimodal. Max score: 100. Categories: instruction_following, multimodal, vision. Language: en. Verified by llm-stats: no.

Leaderboard

  1. Qwen3 VL 235B A22B Thinking self-reported llm-stats
    92.7%