HALC introduces a novel decoding algorithm to reduce object hallucinations in large vision-language models by leveraging fine-grained visual information and a specialized beam search approach.
HALC is a novel decoding algorithm designed to reduce object hallucinations in large vision-language models by integrating adaptive focal-contrast grounding and specialized beam search.
HALC is a novel decoding algorithm designed to reduce object hallucinations in large vision-language models by integrating adaptive focal-contrast grounding and specialized beam search.