cs-cv/vision-language
IB-arxiv-2607.28375-fakett-accuracy
measured
Under the FactGuard temporal protocol, HyperClaim achieved 82.0% accuracy on the FakeTT video-misinformation benchmark.
Observed quantities
accuracy_pct 82.0
01
Primary source
- Title
- HyperClaim: Fine-Grained Cross-Modal Hypergraph Reasoning for Video Misinformation Detection
- Authors
- Xiangbo Wang, Jiasheng Zhang, Xingtong Yu, Luoqiang Lei, Delvin Ce Zhang
- Version
- v1
- Check
- Support quote found word for word in the arXiv abstract v1, 2026-10-08
Under the FactGuard temporal protocol, it achieves 83.7%, 82.0%, and 87.3% accuracy on FakeSV, FakeTT, and FakeVV, respectively
02
Boundary of this block
No additional public limitation was recorded.
03
Connections
Outgoing
No outgoing public edges.
Incoming
No incoming public edges.