基于大模型的证券交易与清算的智能辅助系统及方法
By introducing a legal form-aware restricted decoder and a joint intent encoding mechanism into a large-scale intelligent auxiliary system for securities trading and clearing, the contradiction between form and semantics in the generation of trading instructions is resolved, strong consistency in the execution of trading and clearing is achieved, default risk is reduced, and compliance requirements of financial regulation are met.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- SHANGHAI GUANKAI SOFTWARE TECH CO LTD
- Filing Date
- 2026-03-06
- Publication Date
- 2026-07-17
AI Technical Summary
Existing intelligent auxiliary systems for securities trading and clearing based on large models struggle to perceive and embed the rigid constraints of the clearing process in real time when generating trading instructions. This results in automated instructions being frequently rejected due to format inconsistencies or logical violations, increasing the risk of settlement defaults. Furthermore, the system's reliance on extensive manual verification restricts the reliability and security of intelligent methods in core financial business scenarios.
By introducing a legally form-aware restricted decoder and a joint intent encoding mechanism, compliant trading instructions that meet regulatory and exchange formal requirements are generated. Combined with regulatory rule-anchored inference chains and gradient coordination mechanisms, model parameters are optimized to ensure the formal validity and semantic rationality of the instructions.
It achieves strong consistency in the execution of trading and clearing instructions, reduces the risk of settlement default, ensures the compliance and executability of instructions, meets the requirements of financial regulators for the accuracy and atomicity of instructions, and realizes atomic execution and audit traceability under intelligent decision-making.
Smart Images
Figure CN121810403B_ABST