GenExam: A Multidisciplinary Text-to-Image Exam
Zhaokai Wang, Penghao Yin, Xiangyu Zhao, Changyao Tian et al.
54.40/100
🫥 Mediocre
Incremental, thin
Content 54.4 · Citation bonus +0.0 · 12 citations
💡 GenExam is the first multidisciplinary text-to-image exam benchmark with 1,000 samples across 10 subjects and a four-level taxonomy of exam-style prompts, which evaluates models' semantic correctness
#文生图评估#考试型基准#多学科评测#开源闭源差距#text-to-image evaluation#exam-style benchmark#multidisciplinary assess#open-closed source gap
Score breakdown
Novelty6.0 / 10
Rigor6.0 / 10
Significance7.0 / 10
Clarity8.0 / 10
Reproducibility8.0 / 10
This tone hasn't been generated yet — roast it again to create it.