ArtifactLens: Hundreds of Labels Are Enough for Artifact Detection with VLMs

2026-02-10

AI Summary

This work addresses the challenge of efficiently detecting artifacts—such as malformed hands or distorted objects—in AI-generated images, a task hindered by the heavy reliance of existing methods on large annotated datasets and their limited adaptability to rapidly evolving generative models and emerging artifact types. To overcome these limitations, the authors propose a few-shot artifact detection framework leveraging pre-trained vision-language models (VLMs). By integrating in-context learning, optimized textual prompts, and a multi-component detection architecture, the method achieves state-of-the-art performance with only a few hundred labeled examples per artifact category. Notably, it is the first approach to be uniformly evaluated across five benchmarks, demonstrating strong generalization across morphological, anatomical, and interaction-based artifacts, as well as broader applicability to general AIGC detection tasks.

Technology Category

• Computer Vision: Generative Adversarial Networks (GANs) for Vision
• Machine Learning: Large Multimodal Models (LMMs)
• Humans and AI: AI for Accessibility

Application Category

• Economics, Online Markets and Human Computation: LLM based quality controls for crowd work
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• Social Networks and Social Media: Generative AI / large language models and their impact on social systems

Abstract

Modern image generators produce strikingly realistic images, where only artifacts like distorted hands or warped objects reveal their synthetic origin. Detecting these artifacts is essential: without detection, we cannot benchmark generators or train reward models to improve them. Current detectors fine-tune VLMs on tens of thousands of labeled images, but this is expensive to repeat whenever generators evolve or new artifact types emerge. We show that pretrained VLMs already encode the knowledge needed to detect artifacts - with the right scaffolding, this capability can be unlocked using only a few hundred labeled examples per artifact category. Our system, ArtifactLens, achieves state-of-the-art on five human artifact benchmarks (the first evaluation across multiple datasets) while requiring orders of magnitude less labeled data. The scaffolding consists of a multi-component architecture with in-context learning and text instruction optimization, with novel improvements to each. Our methods generalize to other artifact types - object morphology, animal anatomy, and entity interactions - and to the distinct task of AIGC detection.

Problem

• artifact detection
• visual language models
• AIGC detection
• few-shot learning
• image generation artifacts

Innovation

• artifact detection
• vision-language models
• in-context learning
• text instruction optimization
• few-shot learning