BloomBench: Bilingual English-Arabic VLM Benchmark Grounded in Bloom's Taxonomy
June 3, 2026
BloomBench evaluates vision-language models across six cognitive levels from Bloom's Taxonomy — Remember through Create — using image-question-answer tasks in both English and Arabic. Built with a semi-automated pipeline, it targets gaps in existing VLM evaluations that miss structured cognitive reasoning assessment.
HOW THIS AFFECTS YOU
●
researcherBloomBench provides a structured diagnostic tool for identifying specific cognitive weaknesses in VLMs, particularly useful for Arabic-language model evaluation where benchmarks are scarce.