Multi-Sourced, Multi-Agent Evidence Retrieval for Fact-Checking

Researchers propose WKGFC, a fact-checking system combining knowledge graphs with multi-agent AI reasoning. It addresses misinformation by using authorized knowledge bases as primary information sources, supplemented with web content retrieval. An LLM-powered agent employs Markov Decision Processes to evaluate claims and extract relevant knowledge subgraphs, improving accuracy by capturing multi-hop semantic relationships that traditional text-similarity methods miss.