Back to AI Research

AI Research

JD Oxygen AI Item Center (Oxygen AIIC) V1: An Indus... | AI Research

Key Takeaways

  • JD Oxygen AI Item Center (Oxygen AIIC) V1: An Industrial-Scale LLM/VLM-Centric Solution for Item Understanding, Management, and Applications introduces a lar...
  • this http URL , one of the world's largest e-commerce platforms, serves over 700 million active users and millions of merchants, with a catalog of tens of billions of SKUs.
  • To address these challenges, we present the JD Oxygen AI Item Center (Oxygen AIIC), an industrial-scale platform built on LLMs/VLMs for item-knowledge production and service.
  • Oxygen AIIC now covers tens of thousands of JD categories and processes hundreds of millions of item updates per day on Huawei Ascend NPUs.
  • It has accumulated hundreds of billions of item-knowledge assets.
Paper AbstractExpand

this http URL , one of the world's largest e-commerce platforms, serves over 700 million active users and millions of merchants, with a catalog of tens of billions of SKUs. At this scale, high-quality, structured item knowledge underpins a better consumer experience, lower management costs, and higher operational efficiency-yet producing and serving it poses three industrial-scale challenges: fast-emerging concepts, high-quality knowledge production for massive SKUs, and diverse downstream requirements. To address these challenges, we present the JD Oxygen AI Item Center (Oxygen AIIC), an industrial-scale platform built on LLMs/VLMs for item-knowledge production and service. Oxygen AIIC is built around four core pillars: (i) ontology engineering driven by efficient human-AI collaboration, which supports the dynamic evolution and agile expansion of an ontology with millions of entries; (ii) a "Semantic Search then Discrimination"(S2D) knowledge identification architecture that, combined with throughput improvement strategies, enables scalable, extensible, and high-throughput AI Item Library production for tens of billions of SKUs; (iii) self-evolving item-understanding LLMs/VLMs that improve in a stable and controllable manner, enabling knowledge production with 94.2% precision and 82.8% recall; and (iv) a unified item tunnel that serves as the data and service hub. Oxygen AIIC now covers tens of thousands of JD categories and processes hundreds of millions of item updates per day on Huawei Ascend NPUs. It has accumulated hundreds of billions of item-knowledge assets. Deployed across core business scenarios-including search, recommendation, operations, category planning-Oxygen AIIC has delivered measurable gains at scale. Search-traffic coverage reaches 80.4%, item-information quality issues drop by 37%, the automated fill rate of core attributes during item listing exceeds 80%.

JD Oxygen AI Item Center (Oxygen AIIC) V1: An Industrial-Scale LLM/VLM-Centric Solution for Item Understanding, Management, and Applications introduces a large-scale platform designed to manage the complex data needs of JD.com. As one of the world’s largest e-commerce platforms, JD.com manages tens of billions of items, making it difficult to maintain accurate, structured information. Oxygen AIIC uses Large Language Models (LLMs) and Vision-Language Models (VLMs) to automate the creation and management of item knowledge, helping the platform improve search accuracy, recommendation quality, and operational efficiency.

How Oxygen AIIC Works

The platform is built on four core pillars that work together to manage the entire lifecycle of item data:

  • Ontology Engineering: A human-AI collaborative system where experts define the foundational structure of the product catalog, while AI models dynamically discover and add new concepts as market trends evolve.

  • Knowledge Identification: Using a "Semantic Search then Discrimination" (S2D) architecture, the system efficiently identifies and validates item attributes across billions of products, ensuring the process is scalable and cost-effective.

  • Self-Evolving Models: The platform utilizes LLMs and VLMs that continuously improve through incremental learning, allowing the system to adapt to new data while maintaining high precision and recall.

  • Unified Item Tunnel: A centralized data hub that serves as the backbone for all business applications, providing consistent, high-quality information to various departments, from search engines to category planners.

Scaling to Billions of Items

Managing data for tens of billions of SKUs presents significant technical hurdles, including high computational costs and the risk of AI "hallucinations" (generating incorrect information). To address this, Oxygen AIIC decouples the ontology from the model parameters. This means the system doesn't have to retrain the entire model every time a new product category or attribute is added. By using Huawei Ascend NPUs for high-throughput processing, the platform can handle hundreds of millions of item updates every day, keeping the product catalog fresh and accurate.

Real-World Impact

The deployment of Oxygen AIIC has led to significant improvements in JD.com’s business operations. By automating the filling of product attributes, the platform has achieved an automated fill rate of over 80% for core information. This has resulted in a 37% reduction in item-information quality issues and a 9% increase in click-through rates for optimized product creatives. Furthermore, the system has successfully streamlined category planning, reducing the time required for decision-making from weeks to days.

Key Performance Metrics

The platform demonstrates high reliability in its automated knowledge production, achieving 94.2% precision and 82.8% recall. By integrating these AI-driven insights into core business scenarios, Oxygen AIIC now covers 80.4% of search traffic, proving that large-scale LLM/VLM deployment can effectively bridge the gap between complex e-commerce data and the need for high-quality, structured information.

Comments (0)

No comments yet

Be the first to share your thoughts!