Researchers tackle AI fact-checking failures with new LLM training technique PorBrayan Zanelatto agosto 16, 2024