Title:

AI Agent for Misinformation Detection

This project presents an AI-driven fake news detection system that combines large language model (LLM) reasoning with external credibility signals from trusted sources. The system classifies news articles as Real, Fake, or Unable to Determine using a structured decision process. It integrates three key inputs: source trust scoring, retrieval-augmented context from trusted outlets, and a linguistic flag analysis that detects misinformation patterns and stores results using embeddings. A Gemini-based agent synthesizes these signals to produce consistent, explainable JSON outputs. These outputs return reasonings for the classification of an article. The system also maintains a dynamic memory that updates source credibility over time based on prediction accuracy, enabling adaptive learning. This hybrid approach improves interpretability and robustness compared to standalone LLM classification.

Name

Comment*

Comments are viewable only by submitter

Captcha*