ELIQ: A Label-Free Framework for Quality Assessment of Evolving AI-Generated Images

TL;DR

ELIQ is a novel label-free framework that assesses the quality of evolving AI-generated images by automatically constructing training pairs and adapting a multimodal model for quality prediction.

Contribution

It introduces a transferable, label-free quality assessment method that adapts pre-trained models to evolving AI-generated images without human annotations.

Findings

ELIQ outperforms existing label-free quality assessment methods.

It generalizes well from AI-generated to user-generated content.

ELIQ demonstrates consistent performance across multiple benchmarks.

Abstract

Generative text-to-image models are advancing at an unprecedented pace, continuously shifting the perceptual quality ceiling and rendering previously collected labels unreliable for newer generations. To address this, we present ELIQ, a Label-free Framework for Quality Assessment of Evolving AI-generated Images. Specifically, ELIQ focuses on visual quality and prompt-image alignment, automatically constructs positive and aspect-specific negative pairs to cover both conventional distortions and AIGC-specific distortion modes, enabling transferable supervision without human annotations. Building on these pairs, ELIQ adapts a pre-trained multimodal model into a quality-aware critic via instruction tuning and predicts two-dimensional quality using lightweight gated fusion and a Quality Query Transformer. Experiments across multiple benchmarks demonstrate that ELIQ consistently outperforms existing label-free methods, generalizes from AI-generated content (AIGC) to user-generated content (UGC) scenarios without modification, and paves the way for scalable and label-free quality assessment under continuously evolving generative models. The code will be released upon publication.