China Securities Co., Ltd. has deeply integrated AI technology into its financial business, and independently developed the OctoX—an Intelligent FICC Integration Platform. Focused on the full spectrum of bond products, OctoX efficiently supports end-to-end workflows—from primary market deal sourcing, deal execution, and underwriting to secondary market research, advisory, investment, and market-making—while providing comprehensive, end-to-end risk and compliance management. The platform delivers an intelligent upgrade across bond issuance, investment research and decision-making, transaction execution, and risk control, comprehensively enhancing the FICC business efficiency and empowering high-quality development of the real economy.
In the process of advancing the platform’s development, China Securities achieved an innovative achievement in AI interpretability. The paper “ConciseExplain: Reducing Redundancy and Spuriousness in Persuasive Recommendation Explanation” was formally accepted by the 31st ACM SIGKDD Conference on Knowledge Discovery and Data Mining (KDD 2025). ACM SIGKDD is widely recognized as a premier venue in data mining and knowledge discovery, with rigorous paper acceptance criteria, and is ranked as a Class A conference by the China Computer Federation (CCF).
This research conducts an in-depth investigation into interpretability for bond recommendation scenarios, and innovatively addresses two pervasive shortcomings in widely used explanation methods (e.g., LIME, SHAP, and Integrated Gradients): feature redundancy and spurious correlations. These deficiencies can mislead user decisions and significantly weaken the persuasive power of recommendation systems. Addressing how to construct high-quality recommendation explanations, the project team proposes ConciseExplain, an innovative method that optimizes feature selection and institutes a correlation-validation process, thereby improving the precision and credibility of explanatory outputs.
ConciseExplain not only enriches and strengthens existing approaches in explainable AI (XAI) but also marks a breakthrough in credible explanations for intelligent recommendations in the financial field. It provides a robust technical foundation for building transparent, trustworthy, and user-comprehensible intelligent financial decision-making systems, and holds milestone significance for promoting the development of fintech towards a higher level of intelligence and credibility.
The acceptance of China Securities’ paper by KDD marks the “first-ever breakthrough” for the Chinese securities industry in AI interpretability achievements at CCF A-Class top international academic conferences. It signifies China Securities’ entry into the forefront of international research in the field of artificial intelligence and represents concrete progress in building a closed loop of ” foundational research—technological breakthrough—achievement transformation.”
Looking ahead, under the guidance of the CITIC Group’s FinTech Innovation Center, China Securities will focus on two major directions: enriching the diversity of explanatory expressions and strengthening cross-product extensibility. The goal is to advance explanations from “readable” to “intelligible and actionable,” accelerate deployment across additional business scenarios, and actively align with regulatory standards and industry norms. China Securities will continue to consolidate its leadership in trustworthy AI and intelligent recommendation, injecting sustained momentum into fintech innovation and the company’s high-quality business development.
