SHALE is a synthetic, scalable benchmark that evaluates faithful and factual hallucination in large vision-language models across 12 visual and 6 factual categories, in clean and perturbed settings.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CV 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
SHALE: A Scalable Benchmark for Fine-grained Hallucination Evaluation in LVLMs
SHALE is a synthetic, scalable benchmark that evaluates faithful and factual hallucination in large vision-language models across 12 visual and 6 factual categories, in clean and perturbed settings.