| 标题 |
IMFND: In-context multimodal fake news detection with large visual-language models IMFND:基于大型视觉语言模型的上下文多模态假新闻检测
相关领域
计算机科学
背景(考古学)
视觉语言
自然语言处理
人工智能
语言学
历史
哲学
考古
|
| 网址 | |
| DOI | |
| 其它 | Despite recent advancements, studies indicate that large language models (LLMs), such as GPT-3.5-turbo, often perform worse than well-trained small models, such as BERT, in fake news detection (FND). Similarly, while large visual-language models (LVLMs) demonstrate exceptional capabilities in visual-language reasoning across diverse cross-modal benchmarks, their performance in FND remains unexplored. This paper examines the FND capabilities of LVLMs in comparison to a small but adeptly trained contrastive language-image pre-training (CLIP) model in a zero-shot setting. The results reveal that LVLMs achieve performance comparable to small models. Building on this, the study further explores the integration of standard in-context learning (ICL) with LVLMs, observing modest yet inconsistent improvements in FND performance. To address these limitations, this paper introduces the In-context Multimodal Fake News Detection (IMFND) framework, |
| 求助人 | |
| 下载 | 该求助完结已超 24 小时,文件已从服务器自动删除,无法下载。 |
|
温馨提示:该文献已被科研通 学术中心 收录,前往查看
科研通『学术中心』是文献索引库,收集文献的基本信息(如标题、摘要、期刊、作者、被引量等),不提供下载功能。如需下载文献全文,请通过文献求助获取。
|
PDF的下载单位、IP信息已删除
(2025-6-4)