OmniZip: Learning a Unified and Lightweight Lossless Compressor for Multi-Modal Data

2026-02-25

AI Summary

This work proposes OmniZip, the first lightweight unified framework for lossless compression of multimodal data, addressing the limitation of existing learning-based methods that are typically confined to a single modality and ill-suited for resource-constrained edge devices. OmniZip achieves high compression efficiency across diverse modalities—including images, text, speech, tactile signals, database tables, and genomic sequences—through a modality-agnostic invertible tokenizer, a routing-based contextual modeling and feedforward mechanism, and a reparameterization-aware training strategy, all operating under minimal computational overhead. Experimental results demonstrate that OmniZip outperforms gzip by 42%–62% in compression ratio across multiple benchmarks such as CLIC-M, TouchandGo, enwik9, LibriSpeech, and WikiSQL, while enabling near real-time inference on commodity hardware, including MacBook CPUs and iPhone NPUs, at approximately 1 MB/s.

Technology Category

• Machine Learning: Learning on the Edge & Model Compression
• Computer Vision: Multi-modal Vision
• Data Mining & Knowledge Management: Data Compression

Application Category

• Graph Algorithms and Modeling for the Web: Efficient manipulation of static and dynamic Web-related graphs
• Web Mining and Content Analysis: Mining multimedia, multimodal, multilingual, cross-lingual Web data
• User Modeling, Personalization and Recommendation: On-Device user modeling, personalization, and recommendation

Abstract

Lossless compression is essential for efficient data storage and transmission. Although learning-based lossless compressors achieve strong results, most of them are designed for a single modality, leading to redundant compressor deployments in multi-modal settings. Designing a unified multi-modal compressor is critical yet challenging, as different data types vary largely in format, dimension, and statistics. Multi-modal large language models offer a promising resolution but remain too complex for practical use. Thus, we propose \textbf{OmniZip}, \textbf{a unified and lightweight lossless compressor for multi-modal data (like image, text, speech, tactile, database, and gene sequence)}. Built on a lightweight backbone, OmniZip incorporates three key components to enable efficient multi-modal lossless compression: a modality-unified tokenizer that reversibly transforms diverse data into tokens, a modality-routing context learning mechanism that enables flexible multi-modal context modeling, and a modality-routing feedforward design that further enhances the model's nonlinear representation flexibility. A reparameterization training strategy is used to enhance model capacity. OmniZip outperforms or matches other state-of-the-art compressors on multiple modalities, achieving 42\%, 57\%, 62\% and 42\%, 53\% higher compression efficiency than gzip on CLIC-M, TouchandGo, enwik9, LibriSpeech, and WikiSQL datasets, respectively. It also supports near real-time inference on resource-constrained edge devices, reaching about 1MB/s on MacBook CPUs and iPhone NPUs. Our code is released at https://github.com/adminasmi/OmniZip-CVPR2026.

Problem

• lossless compression
• multi-modal data
• unified compressor
• lightweight model
• data redundancy

Innovation

• unified lossless compression
• multi-modal data
• modality-unified tokenizer
• modality-routing context learning
• lightweight compressor