Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Agentic Multi-Persona Framework for Evidence-Aware Fake News Detection

Roopa Bukke, Soumya Pandey, Suraj Kumar, Soumi Chattopadhyay, Chandranath Adak

TL;DR

<3-5 sentence high-level summary> AMPEND-LS tackles multimodal fake-news detection by integrating structured evidence retrieval (text, images, and knowledge graphs) with an agentic, multi-persona reasoning framework that iteratively validates claims. It combines an LLM-driven reasoning backbone with a lightweight SLM classifier, augmented by persuasion-based refinement to handle uncertainty, and a credibility fusion mechanism that weights evidence by lexical/semantic similarity, domain trust, and temporal relevance. Empirical results across PolitiFact, GossipCop, and MMCoVaR show state-of-the-art accuracy and F1, with robust ablations illustrating the complementary value of each component and positive information-theoretic signals from LLM justifications. The framework offers transparent, scalable decision-making suitable for deployment in dynamic, real-world misinformation contexts, with clear avenues for multilingual and adversarial-robust extensions.

Abstract

The rapid proliferation of online misinformation poses significant risks to public trust, policy, and safety, necessitating reliable automated fake news detection. Existing methods often struggle with multimodal content, domain generalization, and explainability. We propose AMPEND-LS, an agentic multi-persona evidence-grounded framework with LLM-SLM synergy for multimodal fake news detection. AMPEND-LS integrates textual, visual, and contextual signals through a structured reasoning pipeline powered by LLMs, augmented with reverse image search, knowledge graph paths, and persuasion strategy analysis. To improve reliability, we introduce a credibility fusion mechanism combining semantic similarity, domain trustworthiness, and temporal context, and a complementary SLM classifier to mitigate LLM uncertainty and hallucinations. Extensive experiments across three benchmark datasets demonstrate that AMPEND-LS consistently outperformed state-of-the-art baselines in accuracy, F1 score, and robustness. Qualitative case studies further highlight its transparent reasoning and resilience against evolving misinformation. This work advances the development of adaptive, explainable, and evidence-aware systems for safeguarding online information integrity.

Agentic Multi-Persona Framework for Evidence-Aware Fake News Detection

TL;DR

<3-5 sentence high-level summary> AMPEND-LS tackles multimodal fake-news detection by integrating structured evidence retrieval (text, images, and knowledge graphs) with an agentic, multi-persona reasoning framework that iteratively validates claims. It combines an LLM-driven reasoning backbone with a lightweight SLM classifier, augmented by persuasion-based refinement to handle uncertainty, and a credibility fusion mechanism that weights evidence by lexical/semantic similarity, domain trust, and temporal relevance. Empirical results across PolitiFact, GossipCop, and MMCoVaR show state-of-the-art accuracy and F1, with robust ablations illustrating the complementary value of each component and positive information-theoretic signals from LLM justifications. The framework offers transparent, scalable decision-making suitable for deployment in dynamic, real-world misinformation contexts, with clear avenues for multilingual and adversarial-robust extensions.

Abstract

The rapid proliferation of online misinformation poses significant risks to public trust, policy, and safety, necessitating reliable automated fake news detection. Existing methods often struggle with multimodal content, domain generalization, and explainability. We propose AMPEND-LS, an agentic multi-persona evidence-grounded framework with LLM-SLM synergy for multimodal fake news detection. AMPEND-LS integrates textual, visual, and contextual signals through a structured reasoning pipeline powered by LLMs, augmented with reverse image search, knowledge graph paths, and persuasion strategy analysis. To improve reliability, we introduce a credibility fusion mechanism combining semantic similarity, domain trustworthiness, and temporal context, and a complementary SLM classifier to mitigate LLM uncertainty and hallucinations. Extensive experiments across three benchmark datasets demonstrate that AMPEND-LS consistently outperformed state-of-the-art baselines in accuracy, F1 score, and robustness. Qualitative case studies further highlight its transparent reasoning and resilience against evolving misinformation. This work advances the development of adaptive, explainable, and evidence-aware systems for safeguarding online information integrity.
Paper Structure (25 sections, 14 equations, 2 figures, 8 tables, 1 algorithm)

This paper contains 25 sections, 14 equations, 2 figures, 8 tables, 1 algorithm.

Table of Contents

  1. Introduction
  2. Related Work
  3. Problem Formulation
  4. Methodology
  5. Preprocessing
  6. Evidence and Facts Retrieval
  7. Multi-step Multi-persona Agentic Framework (MMAF)
  8. LLM Pseudo-Labeling: Policy, Decision Rules, and Output
  9. Persuasion-based Refinement
  10. LLM Pseudo-Labeling Policy
  11. SLM Classifier for Final Output
  12. Model Deployment
  13. Experiments
  14. Datasets and Evaluation Metrics
  15. Results and Analysis
  16. ...and 10 more sections

Figures (2)

  • Figure 1: Proposed Framework: AMPEND-LS
  • Figure 2: Performance across multi-rounds Q&A on PoltiFact.

Theorems & Definitions (2)

  • definition 1: MMAF
  • definition 2: PKM Agent