LiveFact: A Dynamic, Time-Aware Benchmark for LLM-Driven Fake News Detection

Researchers have developed LiveFact, a new dynamic benchmark for evaluating Large Language Models' ability to detect fake news and misinformation in real-time conditions. The benchmark addresses limitations of static testing by using temporal evidence sets and finds that open-source models like Qwen3-235B-A22B now match proprietary systems in performance.

• →LiveFact introduces a continuously updated benchmark that simulates real-world misinformation detection challenges with temporal uncertainty.
• →Open-source Mixture-of-Experts models now match or outperform proprietary state-of-the-art systems in fake news detection.
• →The benchmark reveals a significant 'reasoning gap' where capable models show epistemic humility by recognizing unverifiable claims.
• →Traditional static benchmarks are vulnerable to data contamination and fail to assess temporal reasoning capabilities.
• →The dual-mode evaluation includes Classification Mode for verification and Inference Mode for evidence-based reasoning.