MUMBAI, India, Aug. 17 -- Intellectual Property India has published a patent application (202641093025 A) filed by Chennai Institute Of Technology on July 31, 2026, for Anti-Hallucination Intelligence Layer For Llm.
Inventors include Dr. A. R. Kavitha; S. Suganya; D. Lita Pansy; S. Mahalakshmi; R. Revathi; and Yuvasri D.
The application for the patent was published on August 14, 2026, under issue no. 33/2026.
Abstract: Title: ANTI-HALLUCINATION INTELLIGENCE LAYER FOR LLM The titled invention discloses an anti-hallucination intelligence system for Large 5 Language Models (LLMs) configured to detect, verify, and suppress hallucinated content before delivering responses to users. The system comprises a Primary Language Model (LLM) (1), a Confidence Entropy Score (CES) Module (2), an Al-to- Al Verification Engine (3), a Decision and Suppression Engine (4), and a Response Assembly Module (5). The Primary Language Model (LLM) (1) generates sentence- 1.0 wise draft responses, while the Confidence Entropy Score (CES) Module (2) evaluates sentence-level uncertainty using internal model confidence signals. The Al-to-Al Verification Engine (3) validates medium- and low-confidence content, and the Decision and Suppression Engine (4) approves, refines, or suppresses unreliable responses before forwarding them to the Response Assembly Module (5). The 15 invention provides a real-time, model-agnostic middleware framework that improves the safety, reliability, and trustworthiness of Al-generated responses by reducing hallucinated content prior to user delivery. NAGAR
Disclaimer: Curated by HT Syndication.
