Starve to Perceive: Taming Lazy Perception in VLMs with Constrained Visual Bandwidth
Open paper
Benchmark
Model
Rank
Results
visual-question-answering-on-v-bench
Starve to Perceive
–
Accuracy: 85.86
visual-question-answering-vqa-on-realworldqa
Starve to Perceive
–
Accuracy: 67.71
Rank counts only results with a code link.