Accelerating Multimodal Large Language Models with Prior-Corrected Token Reduction
Open paper
Benchmark
Model
Rank
Results
object-hallucination-evaluation-on-pope
PriorTR (InternVL2.5-8B, 22.2% tokens)
#1
F1: 90.8
visual-question-answering-vqa-on-seed-seedbench
PriorTR (LLaVA-1.5-7B, 192 tokens)
#27
Accuracy: 64.8
Rank counts only results with a code link.