MUMBAI, India, June 30 -- Intellectual Property India has published a patent application (202641073634 A) filed by R. M. K. College Of Engineering And Technology on June 13, 2026, for Dynamic Context Management Framework For Large Language Model Applications.
Inventors include Dr. Rejin Paul N R; and Dr. M. Arun Manicka Raja.
The application for the patent was published on June 26, 2026, under issue no. 26/2026.
Abstract: The present invention discloses a Dynamic Context Management Framework for Large Language Model applications. The framework intelligently acquires, classifies, stores, prioritizes, compresses, retrieves, and assembles contextual information from multiple sources including user interactions, databases, documents, knowledge repositories, APIs, and real-time data streams. Contextual information is organized into short-term, long-term, semantic, procedural, task-specific, and user-preference memories. A relevance-driven prioritization engine and semantic retrieval mechanism dynamically select context required for inference. A context compression module minimizes token utilization through summarization and knowledge distillation techniques. An adaptive learning engine continuously improves memory allocation, retrieval effectiveness, and personalization using feedback-driven optimization algorithms. The framework enhances response accuracy, reasoning consistency, scalability, personalization, and computational efficiency while reducing hallucinations and inference costs. The invention is particularly applicable to conversational AI, enterprise assistants, autonomous agents, decision-support systems, and intelligent knowledge management platforms requiring long-term contextual awareness and adaptive reasoning capabilities.
Disclaimer: Curated by HT Syndication.
