UniPercept: Towards Unified Perceptual-Level Image Understanding across Aesthetics, Quality, Structure, and Texture

2025-12-25

AI Summary

This work addresses the limited capability of multimodal large language models (MLLMs) in perception-level image understanding—e.g., aesthetics, quality, structure, and texture—by proposing UniPercept, the first unified perceptual understanding framework. Methodologically, it formally defines perception-level understanding tasks; constructs UniPercept-Bench, a large-scale benchmark spanning four core perceptual dimensions; and introduces a joint training paradigm integrating domain-adaptive pretraining with task-aligned reinforcement learning, incorporating hierarchical modeling, dual VQA/VR interfaces, and a plug-and-play reward architecture. Contributions include: (1) substantial performance gains over state-of-the-art MLLMs on perception-centric tasks; (2) seamless integration as a fine-grained perceptual reward module for text-to-image generation models; and (3) establishment of a new benchmark and strong baseline for multimodal perceptual understanding.

Technology Category

• Machine Learning: Large Multimodal Models (LMMs)
• Computer Vision: Large Vision Models
• Natural Language Processing: (Large) Language Models

Application Category

• User Modeling, Personalization and Recommendation: Large Language Models (LLM) for user modeling and recommendation
• Search and Retrieval-Augmented AI: Large language models for search
• Economics, Online Markets and Human Computation: LLM based quality controls for crowd work

Abstract

Multimodal large language models (MLLMs) have achieved remarkable progress in visual understanding tasks such as visual grounding, segmentation, and captioning. However, their ability to perceive perceptual-level image features remains limited. In this work, we present UniPercept-Bench, a unified framework for perceptual-level image understanding across three key domains: Aesthetics, Quality, Structure and Texture. We establish a hierarchical definition system and construct large-scale datasets to evaluate perceptual-level image understanding. Based on this foundation, we develop a strong baseline UniPercept trained via Domain-Adaptive Pre-Training and Task-Aligned RL, enabling robust generalization across both Visual Rating (VR) and Visual Question Answering (VQA) tasks. UniPercept outperforms existing MLLMs on perceptual-level image understanding and can serve as a plug-and-play reward model for text-to-image generation. This work defines Perceptual-Level Image Understanding in the era of MLLMs and, through the introduction of a comprehensive benchmark together with a strong baseline, provides a solid foundation for advancing perceptual-level multimodal image understanding.

Problem

• Enhances MLLMs' perception of image aesthetics, quality, structure, and texture.
• Establishes a benchmark and dataset for perceptual-level image understanding evaluation.
• Develops a unified model for visual rating and question answering tasks.

Innovation

• Unified framework for perceptual-level image understanding
• Domain-Adaptive Pre-Training and Task-Aligned RL training
• Plug-and-play reward model for text-to-image generation