ResearchBenchmarksKnowledge-intensive Question Answering on MuSiQueKnowledge-intensive Question Answering on MuSiQueFollow2.9TruthfulnessGPT-5-14.884-10.267-5.65-1.033Sep 30, 2025Evaluation resultsMethodMethodLinksTruthfulnessHallucinationGPT-52025.092.943.3OpenAI o32025.09149TruthRLBackbone=Llama3.3-70B-...Backbone=Llama3.3-70B-Instruct2025.09-0.123.3PromptingBackbone=Llama3.3-70B-...Backbone=Llama3.3-70B-Instruct2025.09-14.237.7