Organizations: 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 · The University of Queensland, Brisbane, Australia
Multimodal large language models (MLLMs) have substantially advanced video misinformation detection through unified multimodal reasoning, but they often rely on fixed-depth inference and place excessive trust in internally generated assumptions, particularly in scenarios where critical evidence is sparse, fragmented, or requires external verification. To address these limitations, we propose FactGuard, an agentic framework for video misinformation detection that formulates verification as an iterative reasoning process built upon MLLMs. FactGuard explicitly assesses task ambiguity and selectively invokes external tools to acquire critical evidence, enabling progressive refinement of reasoning trajectories. To further strengthen this capability, we introduce a two-stage training strategy that combines domain-specific agentic supervised fine-tuning with decision-aware reinforcement learning to optimize tool usage and calibrate risk-sensitive decision making. Extensive experiments on FakeSV, FakeTT, and FakeVV demonstrate FactGuard's state-of-the-art performance and validate its excellent robustness and generalization capacity.
Figures & tables
Figure 1: Comparison of video misinformation detection methods and our proposed FactGuard in terms of (a) explainability analysis and (b) comprehensive performance analysis.
Figure 2: Pipeline of FactGuard . The upper part illustrates the inference-time agentic verification process, where FactGuard assesses uncertainty based on the ambiguity of the input and selectively invokes external tools to acquire additional evidence before refining its reasoning and producing a final decision. The lower part depicts the training pipeline, which combines supervised fine-tuning with decision-aware reinforcement learning to reinforce structured reasoning, calibrated tool usage, and risk-sensitive verification behavior.
Figure 3: Key advantages of FactGuard. (a) MLLM-based methods with enhanced reasoning may induce cross-modal hallucination in ambiguous cases by over-relying on internally generated assumptions, treating them as grounded evidence without acquiring or validating critical supporting information. (b) FactGuard formulates misinformation verification as an uncertainty-aware, tool-assisted decision-making process that adaptively refines its conclusions, enabling reliable verification in open and dynamic environments.
Model
FakeSV
FakeTT
FakeVV
Acc
Prec
Rec
F1
Acc
Prec
Rec
F1
Acc
Prec
Rec
F1
BERT ( Koroteev, 2021 )
65.4
66.0
66.5
66.2
68.7
67.5
67.5
67.5
60.4
57.9
56.8
57.3
TikTec ( Shang et al., 2021 )
64.8
63.2
61.9
62.5
61.1
64.8
64.2
64.5
59.3
59.1
59.5
59.3
FANVM ( Choi & Ko, 2021 )
65.4
66.1
64.3
65.2
68.9
64.7
68.8
67.1
61.9
60.7
60.8
60.8
SV-FEND ( Qi et al., 2023a )
67.1
67.4
66.3
66.8
67.6
72.2
69.0
70.6
70.9
71.4
71.3
71.3
FakingRec ( Bu et al., 2024 )
69.5
69.7
70.4
70.0
71.0
71.9
72.0
72.0
72.1
72.4
71.6
72.0
Table 1: Performance comparison on FakeSV, FakeTT and FakeVV datasets. We highlight the improvements achieved by FactGuard.
Figure 4: Qualitative analysis of model reasoning. Representative reasoning traces under correct predictions show that FactGuard produces more coherent and evidence-grounded reasoning than Qwen2.5-VL and Fact-R1, highlighting improved interpretability.
Cost Ratio
FakeSV
FakeVV
( α:γ )
Precision
Recall
Precision
Recall
1:2
80.8
82.1
83.2
83.7
1:1
82.2
80.6
85.8
82.1
2:1
83.2
75.0
86.6
79.3
Table 2: Cost-sensitive precision and recall achieved by FactGuard on FakeSV and FakeVV under different asymmetric error cost ratios.
Model
FakeSV
FakeTT
Acc
F1
Acc
F1
FactGuard (Ours)
79.3
81.4
75.3
75.2
w/o SFT
73.1
74.7
71.6
68.2
w/o RL
62.0
63.2
67.9
65.7
w/o Rtool
77.7
78.5
74.7
72.9
w/o Rrisk
78.5
78.9
74.9
73.8
Table 3: Ablation study of FactGuard on FakeSV and FakeTT datasets. We report Accuracy (%) and F1-score (%) to evaluate the contribution of each component.
Variant
FakeSV
FakeTT
Acc
F1
Acc
F1
(a) w/. Rollouts = 6
77.3
78.9
72.6
69.4
(b) w/. Rollouts = 8 (Ours)
79.3
81.4
75.3
75.2
(c) w/. Rollouts = 12
79.5
81.5
75.2
74.3
(d) w/. β = 0.02
78.2
79.5
74.6
73.4
(e) w/. β = 0.04 (Ours)
79.3
81.4
75.3
75.2
Table 4: Performance of GRPO variants on FakeSV and FakeTT.
Appendix figures & tables3 assets
Supplementary material from the paper’s appendix.
Appendix
Figure 5: Additional Case Study of FactGuard.
Figure 6: FactGuard Two-Turn Prompting.
Figure 7: GPT-4o reasoning audit prompt. The prompt evaluates whether a model’s reasoning and prediction are supported by explicit and relevant evidence, serving as an external assessment of reasoning faithfulness.
Real-world multimodal misinformation often involves mixed forgery sources, requiring sample-specific detection strategies. Existing tool-augmented methods rely on predefined workflows or inference-time planning, limiting adaptability or increasing inference cost. To address this issue, we introduce MM-VeriAgent, which learns to verify mixed-source multimodal misinformation with tools. We first build MM-VeriTools, a specialized toolkit for misinformation detection agents. By benchmarking various candidate models and methods on the sub-tasks required by mixed-source detection, we select the strongest for textual, visual, and cross-modal forgery analysis and encapsulate them as callable tools with a unified interface. On top of this toolkit, we train the LVLM agent with reinforcement learning to teach it how to use these tools to better solve mixed-source detection. Since many of the tools are specialized models whose online execution at every rollout severely limits RL efficiency, we further introduce Tool-Execution Cache, which pre-executes candidate tool calls and reuses their cached outputs during training. This preserves multi-step rollouts while reducing online tool execution, largely improving the training efficiency.Experiments on MMFakeBench demonstrate substantial accuracy gains over the base model without explicit tool search at inference time. Ablation and efficiency analyses further validate the learned tool-use policy and show that Tool-Execution Cache reduces online tool executions during training.
Peipei Li, Shuhan Xia, Shengyang Liu +2
Beijing University of Posts and Telecommunications · Minzu University of China · NLPR, Institute of automation, Chinese academy of science, Chinese Academy of Sciences
Multimodal video misinformation detection is commonly formulated as a holistic video-understanding task, where the entire video and its associated content are processed and judged in a single pass. However, real-world misinformation often exhibits a sparse and compositional evidence structure: a reliable decision may depend on only a few coupled clues, while most video content contributes limited additional information. Exhaustive multimodal reasoning may therefore introduce substantial redundancy and obscure decisive evidence. This motivates decoupling evidence acquisition from verification: first identifying sparse, decision-relevant clues and then judging veracity based on the acquired evidence. Accordingly, we propose SIEVE, a framework for Sparse Interactive Evidence Verification via Extraction in multimodal video misinformation detection. An evidence-seeking agent actively explores the available multimodal evidence and constructs a compact evidence package, which is then used by a verifier to determine veracity. The agent is trained with supervised evidence-seeking trajectories and an evidence-aware reinforcement learning objective that promotes informative evidence acquisition while discouraging unnecessary or invalid interactions. Experiments on multiple video misinformation benchmarks show that SIEVE consistently outperforms the evaluated baselines and supports reliable verification using compact evidence packages. Moreover, the resulting acquisition process provides an explicit and inspectable evidence trail, improving the transparency and groundedness of multimodal misinformation detection.
Haochen Zhao, Yongxiu Xu, Xinkui Lin +6
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
Video misinformation detection is often approached through global multimodal fusion or free-form multimodal reasoning. Both paradigms can under-represent localized authenticity cues that arise from coupled interactions among query phrases, contextual text, and short temporal spans of frames. Because such interactions are inherently higher-order, pairwise graph formulations are insufficient to capture multi-way cross-modal dependencies, whereas hypergraphs offer a suitable representation for these relations. We propose HyperClaim, a discriminative temporal hypergraph framework for sample-level authenticity classification. Using the title or benchmark-provided paired text as a claim-like query, HyperClaim constructs a sparse heterogeneous hypergraph over query tokens, evidence tokens, and sampled frames; applies confidence-aware filtering and source budgeting to form compact text-frame and short-range temporal evidence units; performs adaptive soft-incidence reasoning with residual text-video calibration; and aggregates textual, visual, and hyperedge states through a discrepancy-aware readout. Without relying on generated rationales or external tool calls, HyperClaim preserves fine-grained cross-modal and temporal structure that global fusion tends to flatten. Under the FactGuard temporal protocol, it achieves 83.7%, 82.0%, and 87.3% accuracy on FakeSV, FakeTT, and FakeVV, respectively, outperforming strong discriminative and reasoning-centric baselines. Learned incidence and attention weights further reveal token- and frame-level structure.
Xiangbo Wang, Jiasheng Zhang, Xingtong Yu +2
School of Communication Engineering, Hangzhou Dianzi University · School of Computer Science and Technology, Xidian University · School of Computing and Information Systems, Singapore Management University +1