BEIJING, Jan. 02, 2026 (GLOBE NEWSWIRE) -- WiMi Hologram Cloud Inc. (NASDAQ: WiMi) ("WiMi" or the "Company"), a leading global Hologram Augmented Reality ("AR") Technology provider, released a breakthrough achievement—a hybrid quantum-classical deep learning technology based on parameter-efficient quantum modules, QB-Net (Quantum Bottleneck Network). This technology achieves a major breakthrough by embedding lightweight quantum computing modules into the classical U-Net deep learning architecture, reducing the number of parameters in the bottleneck layer by up to 30 times while maintaining performance comparable to that of the classical U-Net. This research and development outcome not only demonstrates the cutting-edge potential of hybrid quantum-classical artificial intelligence but also provides a brand-new optimization paradigm for traditional deep learning architectures.
The core advantage of quantum computing lies in its ability to express high-dimensional information through the superposition states of qubits and perform linear operations in exponentially dimensional spaces, endowing it with expressive and transformative capabilities that surpass classical architectures. However, at the current stage, quantum hardware is still unable to support large-scale quantum neural networks or construct complete quantum U-Net or quantum Transformer.
Therefore, WiMi has taken a completely different path: instead of building fully quantized AI models, it constructs quantum enhancement modules.
This concept stems from a key observation: the bottleneck layer of deep networks is essentially a problem of high-density expression of high-dimensional features, while quantum states are naturally suited to express extremely high-dimensional vector spaces.
When a classical network requires tens of thousands of parameters to accomplish a mapping task, a single quantum state can theoretically achieve the same or even higher expressive power with only a few dozen qubits. This means that as long as classical features can be mapped into quantum states and transformed through quantum circuits, it is possible to achieve equivalent capabilities with extremely low parameter counts.
Based on this idea, WiMi designed a pluggable Quantum Bottleneck Module. This module takes minimal parameter count, structural stability, trainability, and the ability to be integrated into classical networks as its core objectives and has been embedded into the classical U-Net, forming QB-Net.
QB-Net retains the overall structure of U-Net, including the encoder, upsampling path, and skip connections. However, at the bottleneck layer position, the traditional multiple convolutional layers are replaced with a quantum feature compression-transformation-reconstruction module. This module consists of three key steps:
The first step is the encoding of classical features into quantum states. The encoding module uses techniques such as linear projection or amplitude encoding to map the classical feature tensor into a compact vector form suitable for entering quantum circuits. The design of the encoding strategy follows two major principles: minimizing the number of qubits as much as possible while preserving the key information of the features without loss.
The second step is feature transformation through quantum circuits, which is the core link of the entire system and the key to parameter efficiency. A traditional convolutional bottleneck layer may contain hundreds of thousands or even millions of parameters, whereas a quantum circuit requires only tens to hundreds of adjustable rotation parameters to achieve equivalent expressive transformation.
WiMi uses parameterized quantum circuits (PQC) and builds a deeply controllable quantum state transformer through layer stacking. The quantum circuit includes entanglement structures to ensure sufficient information flow between qubits, forming higher-dimensional representation capabilities than classical linear transformations.
The third step is decoding the quantum state back into a classical tensor. The results obtained from quantum measurement are reconstructed through a classical integration and correction module and finally returned to the decoding path of the classical U-Net. The features compressed through the quantum bottleneck retain expressive power yet complete the filtering and abstraction of high-dimensional information with an extremely low number of parameters. The entire process can be directly embedded into existing models without modifying the U-Net architecture or changing the training paradigm, achieving true “plug-and-play quantum enhancement”.
The release of WiMi's QB-Net marks a key step forward for our company on the path of quantum AI technology. It not only proves that quantum computing can deliver real value right now but also demonstrates the enormous potential of deep integration between quantum technology and deep learning. In the future, hybrid quantum-classical architectures will no longer be regarded as transitional technologies but will become one of the mainstream forms of AI for a long time to come.
QB-Net represents a brand-new way of thinking: letting quantum computing become the most valuable part of artificial intelligence rather than the entirety. The hybrid deep learning framework based on parameter-efficient quantum modules will bring a new structural optimization paradigm to the global AI industry and provide a completely new performance improvement path for enterprise-level intelligent systems.

About WiMi Hologram Cloud

WiMi Hologram Cloud Inc. (NASDAQ: WiMi) focuses on holographic cloud services, primarily concentrating on professional fields such as in-vehicle AR holographic HUD, 3D holographic pulse LiDAR, head-mounted light field holographic devices, holographic semiconductors, holographic cloud software, holographic car navigation, metaverse holographic AR/VR devices, and metaverse holographic cloud software. It covers multiple aspects of holographic AR technologies, including in-vehicle holographic AR technology, 3D holographic pulse LiDAR technology, holographic vision semiconductor technology, holographic software development, holographic AR virtual advertising technology, holographic AR virtual entertainment technology, holographic ARSDK payment, interactive holographic virtual communication, metaverse holographic AR technology, and metaverse virtual cloud services. WiMi is a comprehensive holographic cloud technology solution provider. For more information, please visit http://ir.wimiar.com .

Translation Disclaimer

The original version of this announcement is the officially authorized and only legally binding version. If there are any inconsistencies or differences in meaning between the Chinese translation and the original version, the original version shall prevail. WiMi Hologram Cloud Inc. and related institutions and individuals make no guarantees regarding the translated version and assume no responsibility for any direct or indirect losses caused by translation inaccuracies.

Investor Inquiries, please contact:

WIMI Hologram Cloud Inc.
Email: [email protected]

ICR, LLC
Robin Yang
Tel: +1 (646) 975-9495
Email: [email protected]