H-POPE benchmark shows LVLMs hallucinate on object existence and more severely on attributes while often failing to ground outputs in visual input.
Microsoft coco: Common objects in context
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CV 1years
2024 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
H-POPE: Hierarchical Polling-based Probing Evaluation of Hallucinations in Large Vision-Language Models
H-POPE benchmark shows LVLMs hallucinate on object existence and more severely on attributes while often failing to ground outputs in visual input.