dorsal/arxiv
View SchemaProbing Multimodal Large Language Models on Cognitive Biases in Chinese Short-Video Misinformation
| Authors | Jen-tse Huang, Chang Chen, Shiyang Lai, Wenxuan Wang, Michelle R. Kaufman, Mark Dredze |
|---|---|
| Categories | |
| ArXiv ID | 2601.06600vv1 |
| URL | https://arxiv.org/abs/2601.06600 |
| License | http://creativecommons.org/licenses/by-nc-nd/4.0/ |
Abstract
Short-video platforms have become major channels for misinformation, where deceptive claims frequently leverage visual experiments and social cues. While Multimodal Large Language Models (MLLMs) have demonstrated impressive reasoning capabilities, their robustness against misinformation entangled with cognitive biases remains under-explored. In this paper, we introduce a comprehensive evaluation framework using a high-quality, manually annotated dataset of 200 short videos spanning four health domains. This dataset provides fine-grained annotations for three deceptive patterns, experimental errors, logical fallacies, and fabricated claims, each verified by evidence such as national standards and academic literature. We evaluate eight frontier MLLMs across five modality settings. Experimental results demonstrate that Gemini-2.5-Pro achieves the highest performance in the multimodal setting with a belief score of 71.5/100, while o3 performs the worst at 35.2. Furthermore, we investigate social cues that induce false beliefs in videos and find that models are susceptible to biases like authoritative channel IDs.
{
"annotation_id": "110167ef-0d2a-49ce-87ec-ff2739d2f8d7",
"date_created": "2026-02-17T05:53:07.776000Z",
"date_modified": "2026-02-17T05:53:07.776000Z",
"file_hash": "ef197be239ddb2aa6654a430d408c3eaca3cc77ab392f9be76311c4b046a4ca1",
"private": false,
"record": {
"abstract": "Short-video platforms have become major channels for misinformation, where deceptive claims frequently leverage visual experiments and social cues. While Multimodal Large Language Models (MLLMs) have demonstrated impressive reasoning capabilities, their robustness against misinformation entangled with cognitive biases remains under-explored. In this paper, we introduce a comprehensive evaluation framework using a high-quality, manually annotated dataset of 200 short videos spanning four health domains. This dataset provides fine-grained annotations for three deceptive patterns, experimental errors, logical fallacies, and fabricated claims, each verified by evidence such as national standards and academic literature. We evaluate eight frontier MLLMs across five modality settings. Experimental results demonstrate that Gemini-2.5-Pro achieves the highest performance in the multimodal setting with a belief score of 71.5/100, while o3 performs the worst at 35.2. Furthermore, we investigate social cues that induce false beliefs in videos and find that models are susceptible to biases like authoritative channel IDs.",
"arxiv_id": "2601.06600",
"authors": [
"Jen-tse Huang",
"Chang Chen",
"Shiyang Lai",
"Wenxuan Wang",
"Michelle R. Kaufman",
"Mark Dredze"
],
"categories": [
"cs.CL"
],
"license": "http://creativecommons.org/licenses/by-nc-nd/4.0/",
"title": "Probing Multimodal Large Language Models on Cognitive Biases in Chinese Short-Video Misinformation",
"url": "https://arxiv.org/abs/2601.06600",
"version": "v1"
},
"schema_id": "dorsal/arxiv",
"source": {
"execution_id": "fc2de659-f190-46e0-b06b-ef8b5390892e",
"id": "arXiv Dataset",
"type": "Model",
"variant": "snapshot-2026-01-17",
"version": "0.1.0"
},
"user_id": 1000002
}