Zero-shot Quantization: A Comprehensive Survey

2025-05-14

AI Summary

Zero-shot quantization (ZSQ) addresses the practical deployment challenge of low-bit deep model quantization without access to real training data—motivated by privacy, security, and regulatory constraints. This work formally defines ZSQ and introduces the first unified taxonomy, categorizing existing approaches into four classes based on data-generation mechanisms: synthetic-data-driven, statistical-modeling-based, gradient-implicit optimization, and generative-prior-guided methods. We present the inaugural comprehensive survey of ZSQ, rigorously identifying fundamental challenges, methodological boundaries, and establishing standardized evaluation benchmarks. Furthermore, we articulate key future research directions—including scalability, generalization across architectures and tasks, and hardware-aware co-optimization. By bridging theoretical foundations with practical design principles, this work provides both a systematic framework and actionable guidance for efficient, privacy-preserving model compression in sensitive application domains.

Technology Category

• Machine Learning: Learning on the Edge & Model Compression
• Data Mining & Knowledge Management: Data Compression
• Constraint Satisfaction and Optimization: Satisfiability Modulo Theories

Application Category

• User Modeling, Personalization and Recommendation: User privacy protection in personalized systems
• Security and Privacy: Large-scale security measurements
• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs

Abstract

Network quantization has proven to be a powerful approach to reduce the memory and computational demands of deep learning models for deployment on resource-constrained devices. However, traditional quantization methods often rely on access to training data, which is impractical in many real-world scenarios due to privacy, security, or regulatory constraints. Zero-shot Quantization (ZSQ) emerges as a promising solution, achieving quantization without requiring any real data. In this paper, we provide a comprehensive overview of ZSQ methods and their recent advancements. First, we provide a formal definition of the ZSQ problem and highlight the key challenges. Then, we categorize the existing ZSQ methods into classes based on data generation strategies, and analyze their motivations, core ideas, and key takeaways. Lastly, we suggest future research directions to address the remaining limitations and advance the field of ZSQ. To the best of our knowledge, this paper is the first in-depth survey on ZSQ.

Problem

• Quantizing deep learning models without training data
• Addressing privacy and security constraints in quantization
• Surveying zero-shot quantization methods and advancements

Innovation

• Zero-shot quantization without real data
• Categorizes methods by data generation strategies
• First in-depth survey on ZSQ techniques