Verus-Insight targets explainable artifact detection in AI-generated video
Original title: Verus-Insight: Towards Explainable Artifact Detection in AI-Generated Videos
The method splits detection into inspection planning, evidence verification, and aggregation, and is evaluated on a new human-annotated benchmark.
Researchers introduce Verus-Insight, an artifact detector for AI-generated video trained with human-anchored full-process supervision and progressive capability training. They frame detection as three steps: deciding what to inspect, verifying genuine generation artifacts, and aggregating that evidence into a source decision. They also release Verus-Bench, with fine-grained human annotations of perceptible artifacts. The paper reports stronger fine-grained artifact recovery and video-level detection than prior artifact-based detectors, without more spurious artifact claims.
Unconfirmed: Reported gains come from the authors' experiments and are not independently confirmed here.
- Verus-Insight uses inspection planning, evidence verification, and evidence aggregation.
- Training uses human-anchored full-process supervision and progressive capability training.
- Verus-Bench supplies fine-grained human annotations of perceptible artifacts.
- Authors say the three capabilities are complementary but distinct.
- They report more reliable conversion of inspection targets into supported artifact evidence.
- AI models
- Verus-Insight
From the original article
1 min read at huggingface.co
Recent advances in video generation have made synthetic videos increasingly realistic, placing greater demands on reliable AI-generated video detection. Existing artifact-based detectors typically identify generation artifacts as evidence for source prediction, yet they may either fail to recognize subtle artifacts or reach the correct source decision without faithfully identifying the evidence underlying it.
By Zhuoran Zhang, Yang Shi, Qixun Wang · huggingface.co
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.