Compositional Zero-Shot Learning: A Survey

2025-10-13

AI Summary

This work addresses the core challenge in compositional zero-shot learning (CZSL): recognizing unseen attribute-object compositions (e.g., “wet cat”) given only seen ones (e.g., “wet car”, “dry cat”), where visual representations exhibit strong contextual dependence. We propose the first decoupling-based CZSL classification framework, systematically categorizing methods into four paradigms: non-explicit decoupling, text-only decoupling, vision-only decoupling, and cross-modal decoupling—distinguishing both closed-set and open-set evaluation settings. Our approach integrates decoupled representation learning, cross-modal alignment, and context-aware semantic embedding, validated through comprehensive analysis on large-scale vision-language datasets. The study provides a unified technical taxonomy, identifies fundamental bottlenecks in modeling context sensitivity and compositional generalization, and releases an open-source benchmark—including standardized evaluation protocols, datasets, and code—to establish both theoretical foundations and practical tools for advancing CZSL research.

Technology Category

• Computer Vision: Representation Learning for Vision
• Machine Learning: Multimodal Learning
• Natural Language Processing: Language Grounding & Multi-modal NLP

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Graph Algorithms and Modeling for the Web: Graph embeddings and representation learning for Web-related graphs
• Semantics and Knowledge: Scalable techniques for the creation, curation, publication, maintenance, and consumption of large, Web-based, structured, reusable, knowledge graphs and ontologies

Abstract

Compositional Zero-Shot Learning (CZSL) is a critical task in computer vision that enables models to recognize unseen combinations of known attributes and objects during inference, addressing the combinatorial challenge of requiring training data for every possible composition. This is particularly challenging because the visual appearance of primitives is highly contextual; for example, ``small'' cats appear visually distinct from ``older'' ones, and ``wet'' cars differ significantly from ``wet'' cats. Effectively modeling this contextuality and the inherent compositionality is crucial for robust compositional zero-shot recognition. This paper presents, to our knowledge, the first comprehensive survey specifically focused on Compositional Zero-Shot Learning. We systematically review the state-of-the-art CZSL methods, introducing a taxonomy grounded in disentanglement, with four families of approaches: no explicit disentanglement, textual disentanglement, visual disentanglement, and cross-modal disentanglement. We provide a detailed comparative analysis of these methods, highlighting their core advantages and limitations in different problem settings, such as closed-world and open-world CZSL. Finally, we identify the most significant open challenges and outline promising future research directions. This survey aims to serve as a foundational resource to guide and inspire further advancements in this fascinating and important field. Papers studied in this survey with their official code are available on our github: https://github.com/ans92/Compositional-Zero-Shot-Learning

Problem

• Surveying compositional zero-shot learning methods for unseen attribute-object recognition
• Addressing contextual visual appearance challenges in primitive combinations
• Providing taxonomy and comparative analysis of disentanglement approaches

Innovation

• Surveying compositional zero-shot learning methods
• Introducing taxonomy based on disentanglement approaches
• Analyzing methods across closed and open-world settings