Unfold The World: Factorize 4D Properties in Reinforcing Spatial Reasoning

Benchmark Model Rank Results
visual-question-answering-vqa-on-mmbench-cnFactoSR-8B-RL#10Accuracy: 84.1
visual-question-answering-vqa-on-mmbench-enFactoSR-8B-RL#13Accuracy: 85.7
optical-character-recognition-ocr-on-ocrbenchFactoSR-8B-SFT#17Score: 877
visual-question-answering-vqa-on-blinkFactoSR-8B-RL#17Accuracy: 66
visual-question-answering-vqa-on-realworldqaFactoSR-8B-RL#41Accuracy: 71.6
vision-language-understanding-on-mmstarFactoSR-8B-RL#44Accuracy: 69.1