ResearchBenchmarksHonesty Alignment on In-Domain Aggregate AverageFollow84.36AUROCEliCal79.347280.648681.9583.2514Oct 20, 2025Evaluation ResultsMethodMethodLinksAUROCEliCalModel=Qwen-7B, Regime=...Model=Qwen-7B, Regime=Training-based, Sample Size=1k2025.1084.36EliCalModel=Qwen-14B, Regime...Model=Qwen-14B, Regime=Training-based, Sample Size=1k2025.1083.3EliCalModel=Llama-8B, Regime...Model=Llama-8B, Regime=Training-based, Sample Size=1k2025.1079.54