From Manipulation to Mistrust: Explaining Diverse Micro-Video Misinformation for Robust Debunking in the Wild

Zhi Zeng; Yifei Yang; Jiaying Wu; Xulang Zhang; Xiangzheng Kong; Herun Wan; Zihan Ma; Minnan Luo

cs.SIMar 26, 2026

From Manipulation to Mistrust: Explaining Diverse Micro-Video Misinformation for Robust Debunking in the Wild

Authors: Zhi Zeng, Yifei Yang, Jiaying Wu, Xulang Zhang, Xiangzheng Kong, Herun Wan, Zihan Ma, Minnan Luo

Organizations: School of Computer Science and Technology, MOEKLINNS Lab, Xi’an Jiaotong University, Xi’an, Shaanxi, China · National University of Singapore, Singapore · Xi’an Jiaotong University, Xi’an, Shaanxi, China

Abstract

The rise of micro-videos has reshaped how misinformation spreads, amplifying its speed, reach, and impact on public trust. Existing benchmarks typically focus on a single deception type, overlooking the diversity of real-world cases that involve multimodal manipulation, AI-generated content, cognitive bias, and out-of-context reuse. Meanwhile, most detection models lack fine-grained attribution, limiting interpretability and practical utility. To address these gaps, we introduce WildFakeBench, a large-scale benchmark of over 10,000 real-world micro-videos covering diverse misinformation types and sources, each annotated with expert-defined attribution labels. Building on this foundation, we develop FakeAgent, a Delphi-inspired multi-agent reasoning framework that integrates multimodal understanding with external evidence for attribution-grounded analysis. FakeAgent jointly analyzes content and retrieved evidence to identify manipulation, recognize cognitive and AI-generated patterns, and detect out-of-context misinformation. Extensive experiments show that FakeAgent consistently outperforms existing MLLMs across all misinformation types, while WildFakeBench provides a realistic and challenging testbed for advancing explainable micro-video misinformation detection. Data and code are available at: https://github.com/Aiyistan/FakeAgent.

Figures & tables

Figure 1. WildFakeBench at a glance: over 10,000 real-world micro-videos capturing diverse forms of misinformation.

Figure 2. Comparison of our proposed FakeAgent, neural-network-based and LLM-based methods.

Datasets	Time Span	#Post (Misinformation/Real)	#Video	Type	In Wild	Source
FVC ( Papadopoulou et al., 2019 )	-2018	2,916/2,090	5,006	1	$\surd$	YT/TW
(Palod et al. 2019) ( Palod et al., 2019 )	2013-2016	123/423	546	1	$\surd$	FB
(Hou et al. 2019) ( Hou et al., 2019 )	-2019	118/132	250	1	$\surd$	TT
(Serrano et al. 2020) ( Serrano et al., 2020 )	-2020	113/67	180	1	$\surd$	YT
(Choi and Ko 2021) ( Choi and Ko, 2021 )	-2021	902/903	1,805	1	$\surd$	YT
(Shang et al. 2021) ( Shang et al., 2021 )	-2020	226/665	891	1	$\surd$	TT

Table 1. Comparison of micro-video misinformation benchmarks. WildFakeBench spans the longest period, includes the most videos, and covers the most diverse deception types and sources. Source platforms: YT (YouTube), TW (Twitter), FB (Facebook), TT (TikTok), BB (Bilibili), WB (Weibo), DY (Douyin), IS (Instagram), KS (Kuaishou).

Figure 3. Overview of the data annotation process.

Panel of Figure 4 — Figure 4. Data analysis of our WildFakeBench.

Panel of Figure 5 — Figure 5. Domain-specific word clouds in WildFakeBench.

Figure 6. Overview of our proposed FakeAgent framework.

	Modality			Content Manipulation			Semantic Biases			Out-of-Context			AIGC	Mean
	A	V	T	TT	VT	AT	FL	EN	OC	KE	EF	ES	AIGC	Mean
Smaller-Parameter MLLMs
Qwen2-Audio-7B (Direct)	$\surd$	$\times$	$\surd$	41.02	39.20	34.33	70.42	43.30	22.92	35.48	40.28	43.88	21.39	49.29
InternVL-3-8B (Direct)	$\times$	$\surd$	$\surd$	14.83	1.78	6.67	29.33	28.87	24.65	24.19	19.14	11.22	36.87	19.76
InternVL-2.5-8B (Direct)	$\times$	$\surd$	$\surd$	14.33	12.00	13.33	13.33	65.94	81.94	64.92	72.53	74.15	91.98	50.45
Qwen2.5-VL-7B (Direct)	$\times$	$\surd$	$\surd$	34.33	9.78	13.33	33.33	70.10	60.62	65.73	62.65	52.38	95.51	49.78

Table 2. Main results (%). Performance of all models on the primary evaluation metric, Micro-Acc. Misinformation types are abbreviated: TT (Text Tampering), VT (Video Tampering), AT (Audio Tampering), FL (Faulty Logic), EN (Exaggerated Narration), OC (Offensive Content), KE (Knowledge Error), EF (Event Fabrication), ES (Event Splicing), and AIGC (AI-Generated Content).

Model	Acc	F1	Macro-P	Macro-R
w/o Text	61.34	54.74	58.16	55.86
w/o Video	61.15	54.49	57.86	55.65
w/o Audio	59.62	53.92	56.01	54.71
w/o CKR	60.49	52.29	56.36	54.16
w/o EER	63.49	61.19	59.87	59.89
FakeAgent-7B	67.98	65.42	66.39	65.68

Table 3. Ablation results (%). Macro-level performance of different model variants on WildFakeBench.

Figure 7. Evaluation of model explanation quality using both automatic and human assessments.

Panel of Figure 8 — Figure 8. Qualitative examples illustrating how FakeAgent detects and explains different types of micro-video misinformation.

Appendix figures & tables1 asset

Supplementary material from the paper’s appendix.

Appendix

Figure 9. Error case. A real-world example highlighting the limitation of FakeAgent in handling domain-specific misinformation.

Fake News Detection

Fake news detection identifies false or misleading claims by analyzing content, context, evidence, and how information spreads.

See all 98 papers

When Seeing Is Not Believing -- A Benchmark for Search-Grounded Video Misinformation Detection
Jun 2, 2026Tao Yu, Yujia Yang, Shenghua Chai +17Fake News Detection Automated Fact-Checking
CASIA · UCAS · BAAI +2
FactGuard: Agentic Video Misinformation Detection via Reinforcement Learning
Feb 26, 2026Zehao Li, Hongwei Yu, Hao Jiang +7Automated Fact-Checking Tool Use in VLMs
Institute of Computing Technology, Chinese Academy of Sciences, Beijing, China · University of Chinese Academy of Sciences, Beijing, China · University of Science and Technology Beijing, Beijing, China +1
Sparse Evidence Can Suffice: Agentic Evidence Seeking for Multimodal Video Misinformation Detection
Jul 20, 2026Haochen Zhao, Yongxiu Xu, Xinkui Lin +6Automated Fact-Checking Multimodal Agents
Institute of Information Engineering, Chinese Academy of Sciences, Beijing, China · School of Cyber Security, University of Chinese Academy of Sciences, Beijing, China · Baidu, Beijing, China

Jun 2, 2026J/K move · Enter open · S save

When Seeing Is Not Believing -- A Benchmark for Search-Grounded Video Misinformation Detection

Tao Yu, Yujia Yang, Shenghua Chai, Jinshuai Zhang+16

CASIA · UCAS · BAAI+2

Fake News Detection Automated Fact-Checking

Open paper AskPDF arXiv

Video misinformation increasingly operates at the semantic and evidential level: authentic footage may be selectively edited, temporally reordered, spliced across sources, or augmented with AI-generated content to construct false narratives. Such evidence-dependent manipulations cannot be reliably verified from the input video alone, because the missing, reordered, replaced, or recontextualized evidence lies outside the video itself. We introduce EVID-Bench, a benchmark for search-grounded video misinformation detection, where a system must search the open web for related videos and identify what information is false through cross-video comparison. EVID-Bench comprises 222 videos spanning 9 manipulation types across 3 categories: AI generation, single-source editing, and multi-source editing. All samples are verified to be undetectable by frontier models through visual inspection alone. We evaluate nine frontier multimodal models using a retrieval-augmented verification baseline. The best system achieves only 61.43% point-level accuracy and 43.24% video-level accuracy, while AI-generated manipulations remain especially challenging. Error analysis reveals recurring challenges: models fixate on irrelevant anchors, misattribute synthetic content to editorial splicing, and terminate search prematurely before fully explaining the manipulation.

Abstract

Fake News Detection

Explore similar work

When Seeing Is Not Believing -- A Benchmark for Search-Grounded Video Misinformation Detection

FactGuard: Agentic Video Misinformation Detection via Reinforcement Learning

Sparse Evidence Can Suffice: Agentic Evidence Seeking for Multimodal Video Misinformation Detection