Mini-Monkey: Alleviating the Semantic Sawtooth Effect for Lightweight MLLMs via Complementary Image Pyramid

Benchmark Model Rank Results
infographic-visual-question-answering-on-infovqaMini-Monkey (extra data)#8ANLS: 60.1
visual-question-answering-vqa-on-textvqa-valMini-Monkey (extra data)#19Accuracy: 76.0
optical-character-recognition-ocr-on-ocrbenchMini-Monkey (extra data)#51Score: 806
visual-question-answering-on-docvqa-testMini-Monkey (extra data)#55ANLS: 0.874
chart-question-answering-on-chartqaMini-Monkey (extra data)#781:1 Accuracy: 76.5