DeLIVeR framework enhances LLM fact-checking with knowledge graph exploration · 1 source tracked

Researchers have developed DeLIVeR, a new framework designed to improve the accuracy of automated fact-checking by large language models. This system decomposes complex claims into targeted questions, which are then used to explore knowledge graphs for evidence. The framework employs a reinforced learning approach, specifically Group Relative Policy Optimization (GRPO), to train a planner LLM, leading to significant improvements in veracity recognition. Evaluations on datasets like LIAR, FEVER, and PolitiFact demonstrated that DeLIVeR, when using Qwen2.5-7B, achieved F1-scores up to 15% higher than existing methods, offering a more transparent and auditable approach to misinformation detection. AI

IMPACT Enhances LLM capabilities in fact-checking and misinformation detection by improving evidence retrieval and reasoning.

RANK_REASON The cluster contains a research paper detailing a new framework for LLM fact-checking. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.AI →

• FEVER
• Group Relative Policy Optimization
• GRPO
• HippoRAG2
• knowledge graph
• large-language models
• PolitiFact
• Qwen2.5-7B

• paper
• model release

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →