Intellectual Property India Publishes Patent Application for System And Method For Agile Task Management Using Llm-Driven Reinforcement And Probabilistic Adversarial Error Detection

MUMBAI, India, Aug. 10 -- Intellectual Property India has published a patent application (202641091568 A) filed by Mrs. N. Durga Devi; and Dr. Tirimula Rao Benala on July 28, 2026, for System And Method For Agile Task Management Using Llm-Driven Reinforcement And Probabilistic Adversarial Error Detection.

Inventors include Mrs. N. Durga Devi; and Dr. Tirimula Rao Benala.

The application for the patent was published on July 31, 2026, under issue no. 31/2026.

Abstract: ABSTRACT SYSTEM AND METHOD FOR AGILE TASK MANAGEMENT USING LLM-DRIVEN REINFORCEMENT AND PROBABILISTIC ADVERSARIAL ERROR DETECTION The present invention discloses a system and method for agile task management with integrated malicious error detection, combining Large Language Models (LLMs) with deep reinforcement Gaussian Bayes adversarial neural networks (DR-GBANN). The invention preprocesses agile task-based source code to normalize syntax, tokenize functional components, and extract task-level semantics. The LLM module generates contextual embeddings representing execution logic, dependencies, and hidden behavioral patterns. These embeddings are analyzed by the DR-GBANN, which integrates reinforcement learning for adaptive policy optimization and probabilistic adversarial inference for robust detection of malicious task behaviors under uncertain execution conditions. The system dynamically adapts to evolving task definitions, rewarding accurate detection decisions and penalizing misclassifications, while probabilistic adversarial reasoning enhances resilience against obfuscated or adversarially manipulated code. Experimental validation demonstrates superior performance compared to conventional rule-based, SVM, and CNN-LSTM models, achieving throughput of 103 tasks per second, reduced delay of 55 milliseconds, optimized resource utilization of 61%, and malicious error detection accuracy of 93%. The invention ensures secure, efficient, and scalable agile task execution, making it suitable for deployment in distributed DevOps environments and continuous integration pipelines.

Disclaimer: Curated by HT Syndication.