arXiv 2512.229332026-05-12cs.AIcs.CL

RW-Post：可审计的证据导向多模态事实核查

RW-Post: Auditable Evidence-Grounded Multimodal Fact-Checking in the Wild

School of Computing (SoC), National University of Singapore (NUS)（新加坡国立大学计算机学院（SoC））；National University of Singapore (NUS)（新加坡国立大学）；Department of Electrical and Computer Engineering (ECE), National University of Singapore (NUS)（新加坡国立大学电子与计算机工程系（ECE））

Danni Xu, Shaojing Fan, Harry Cheng, Mohan Kankanhalli

备注：: Code and dataset will be released at https://github.com/xudanni0927/AgentFact

阅读 PDF

arXiv 页面 ↗HTML 原文 ↗

AI总结：

RW-Post提出一种可审计的多模态事实核查基准，通过人类事实核查文章提取证据，支持不同场景下的可控评估，提升视觉 grounding 和证据利用能力。

摘要：

多模态虚假信息越来越多地利用视觉说服，其中篡改或再利用的图像强化误导性文本。我们介绍了RW-Post，一种与文本-图像对齐的基准，用于真实世界的多模态事实核查，具有可审计的注释：每个实例链接原始社交媒体帖子与推理轨迹和从人类事实核查文章中通过LLM辅助提取和审计流程提取的显式证据项。RW-Post支持在闭书、证据受限和开放网络场景下的受控评估，使系统性诊断视觉 grounding 和证据利用成为可能。我们提供AgentFact作为参考验证基线，并在统一协议下基准强大的开源LVLMs。实验表明存在显著提升空间：当前模型在忠实证据 grounding 方面表现不佳，而证据受限评估同时提高了准确性和忠实性。

英文摘要：

Multimodal misinformation increasingly leverages visual persuasion, where repurposed or manipulated images strengthen misleading text. We introduce RW-Post, a post-aligned text--image benchmark for real-world multimodal fact-checking with auditable annotations: each instance links the original social-media post with reasoning traces and explicitly linked evidence items derived from human fact-check articles via an LLM-assisted extraction-and-auditing pipeline. RW-Post supports controlled evaluation across closed-book, evidence-bounded, and open-web regimes, enabling systematic diagnosis of visual grounding and evidence utilization. We provide AgentFact as a reference verification baseline and benchmark strong open-source LVLMs under unified protocols. Experiments show substantial headroom: current models struggle with faithful evidence grounding, while evidence-bounded evaluation improves both accuracy and faithfulness.

机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。