ALOHa: A New Measure for Hallucination in Captioning Models

arXiv:2404.02904v1 Announce Type: cross
Abstract: Despite recent advances in multimodal pre-training for visual description, state-of-the-art models still produce captions containing errors, such as hallucinating objects not present in a scene. The existing prominent metric for object hallucination, CHAIR, is limited to a fixed set of MS COCO objects and synonyms. In this work, we propose a modernized open-vocabulary metric, ALOHa, which leverages large language models (LLMs) to measure object hallucinations. Specifically, we use an LLM to extract groundable objects from …

aloha arxiv captioning cs.ai cs.cl cs.cv cs.lg hallucination type