MUMBAI, India, Sept. 22 -- Intellectual Property India has published a patent application (202611094356 A) filed by Thapar Institute Of Engineering & Technology on August 04, 2026, for A Context Optimization Framework For Dual Memory Retrieval-Augmented Intelligent Systems.
Inventors include Aditya Midha; Prof. Mukesh Singh; and Dr. Prashant Singh Rana.
The application for the patent was published on September 18, 2026, under issue no. 38/2026.
Abstract: ABSTRACT A CONTEXT OPTIMIZATION FRAMEWORK FOR DUAL MEMORY RETRIEVAL-AUGMENTED INTELLIGENT SYSTEMS The present invention relates to a Context-Orchestrated Dual Memory Architecture (CODMA) for context-optimized retrieval-augmented intelligent systems. The invention provides a retrieval and reasoning framework that performs explicit context optimization before information retrieval and language model inference to improve response accuracy and reduce hallucinations. The architecture comprises a Context Fusion Engine (CTX) configured to generate and optimize a structured five- dimensional contextual representation based on semantic relevance, evidential coherence, memory significance, source trustworthiness, and inferred intent. A Dual Knowledge Registry (DKR) includes a Global Knowledge Registry (GKR) and an Enterprise Knowledge Registry (EKR).. The system further incorporates adaptive retrieval routing, pre-retrieval trust-based enterprise access control, evidence fusion, and reasoning prior to language generation. Context optimization is performed under predefined constraints including token budget, trust threshold, latency, and computational cost, thereby improving retrieval precision, reasoning consistency, enterprise security, computational efficiency, and response reliability across diverse intelligent information retrieval and knowledge management applications.
Disclaimer: Curated by HT Syndication.
