Boosting the Power of Small Multimodal Reasoning Models to Match Larger Models with Self-Consistency Training
Open paper
Benchmark
Model
Rank
Results
science-question-answering-on-scienceqa
MC-CoT F-Large
#1
Avg. Accuracy: 94.88
Natural Science: 97.47
Social Science: 90.44
…
visual-question-answering-on-a-okvqa
MC-CoT
#3
MC Accuracy: 71
Rank counts only results with a code link.