paper_type.json (328B)
1 { 2 "paper_type": "empirical", 3 "reason": "Proposes a framework and reports quantitative experimental results (92% detection accuracy, 65% reduction in deceptive outputs, 78% fairness improvement) across multiple benchmarks, even though the evidence is limited to toy models rather than the claimed enterprise-scale systems." 4 }