ResearchBenchmarksSafety Evaluation on LM Safety Evaluation DatasetFollow0Unsafe RateGoal-0.9725.58912.1518.711Apr 17, 2026Evaluation ResultsMethodMethodLinksUnsafe RateOver-Refusal RateGoalModel=Mistral-NemoModel=Mistral-Nemo2026.04041.7GoalModel=Mistral-SmallModel=Mistral-Small2026.04061.3BeamModel=Mistral-SmallModel=Mistral-Small2026.043.8324.3BeamModel=Mistral-NemoModel=Mistral-Nemo2026.044.3325.3BaseModel=Mistral-SmallModel=Mistral-Small2026.0415.816.3CBModel=Mistral-SmallModel=Mistral-Small2026.0417.216.3BaseModel=Mistral-NemoModel=Mistral-Nemo2026.0424.217.7CBModel=Mistral-NemoModel=Mistral-Nemo2026.0424.315