Unfold The World: Factorize 4D Properties in Reinforcing Spatial Understanding
Open paper
Benchmark
Model
Rank
Results
spatial-reasoning-on-vsi-bench
FactoSR-8B-RL
–
Accuracy: 61.5
visual-question-answering-vqa-on-blink
FactoSR-8B-RL
–
Accuracy: 55.4
visual-question-answering-vqa-on-realworldqa
FactoSR-8B-RL
–
Accuracy: 47.3
Rank counts only results with a code link.