List your company

Content Safety Classification on ToxicChat

75.46PrecisionGradSafe
6.944824.732442.5260.3076Apr 6, 2026

Evaluation results

MethodLinks
GradSafe2026.0475.4666.3970.641.5533.6
Top-pDecoding Strategy=Top-p2026.0473.0757.5764.4110.5142.42
Top-kDecoding Strategy=Top-k2026.0468.4950.558.1411.5149.49
GreedyDecoding Strategy=Greedy2026.0468.3767.6768.0215.532.32
Gradient-Controlled Decoding (GCD)2026.0467.3472.6769.912.5427.32
Safe-Decoding2026.049.5892.6217.3762.897.38