AI Paper Atlas

Research map / Video / Video-language understanding

AVFakeBench: A Comprehensive Audio-Video Forgery Detection Benchmark for AV-LMMs

Shuhan Xia, Peipei Li, Xuannan Liu, Dongsen Zhang, Xinyu Guo, Zekun Li

CVPR 2026 · Video › Video-language understanding · approach: Event and anomaly reasoning in video · 1 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
evaluate on general-subject instead of human-subject clipsLipFD · AVFakeBench Binary JudgmentF1: 45.0 -> 18.9Hurt
evaluate on general-subject instead of human-subject clipsAVH-Align · AVFakeBench Binary JudgmentF1: 51.7 -> 34.3Hurt
evaluate on general-subject instead of human-subject clipsGemini-2.5-pro · AVFakeBench Binary JudgmentF1: 63.3 -> 54.3Hurt
use Gemini-2.5-pro AV-LMM instead of expert detector AVH-AlignGemini-2.5-pro · AVFakeBench Binary Judgment (overall)macro-F1: 42.8 -> 59.9Improved
switch task from binary judgment to forgery type classificationGemini-2.5-pro · AVFakeBenchmacro-F1: 59.9 -> 19.2Hurt
edited forgeries instead of real samples in classificationGemini-2.5-pro · AVFakeBench Forgery ClassificationF1: 42.8 -> 7.5Hurt
edited forgeries instead of synthesized samples in classificationGemini-2.5-pro · AVFakeBench Forgery ClassificationF1: 13.9 -> 7.5Hurt
video-only forgery instead of audio-only forgeryGPT-4o · AVFakeBench Forgery ClassificationF1: 8.0 -> 21.7Improved
video-only forgery instead of audio-only forgeryOneLLM · AVFakeBench Forgery ClassificationF1: 0.0 -> 18.8Improved
video-only forgery instead of audio-only forgeryVideo-LLaMA2 · AVFakeBench Forgery ClassificationF1: 0.0 -> 14.7Improved
audio-only forgery instead of video-only forgeryGemini-2.5-pro · AVFakeBench Forgery ClassificationF1: 23.0 -> 28.4Improved
proprietary GPT-4o instead of best open-source video-SALMONN on detail selectionGPT-4o · AVFakeBench Forgery Detail Selectionmacro-F1: 17.8 -> 27.5Improved

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy