List your company

Safety Evaluation on StrongReject (SR)

16.7Reasoning Harmful RatioIPO
14.81227.55640.353.044Sep 29, 2025

Evaluation results

MethodLinks
IPOBackbone=DeepSeek-R1-Distill-Llama-8B2025.0916.710.9
GRPOBackbone=DeepSeek-R1-Distill-Llama-8B2025.091919.7
IPOBackbone=Qwen3-8B2025.0919.313.7
IPOBackbone=DeepSeek-R1-Distill-Qwen-7B2025.0920.78.4
STARBackbone=DeepSeek-R1-Distill-Llama-8B2025.0921.914.6
GRPOBackbone=Qwen3-8B2025.0923.222.6
STARBackbone=DeepSeek-R1-Distill-Qwen-7B2025.0924.917.8
GRPOBackbone=DeepSeek-R1-Distill-Qwen-7B2025.0925.121.9
SafeKeyBackbone=DeepSeek-R1-Distill-Qwen-7B2025.0931.824.4
SafeKeyBackbone=DeepSeek-R1-Distill-Llama-8B2025.0934.323.3
RealSafeBackbone=DeepSeek-R1-Distill-Llama-8B2025.0934.76.1
RealSafeBackbone=DeepSeek-R1-Distill-Qwen-7B2025.0936.52.1
BaseBackbone=Qwen3-8B2025.0949.633.8
SafeChainBackbone=DeepSeek-R1-Distill-Llama-8B2025.0955.338.7
SafeChainBackbone=DeepSeek-R1-Distill-Qwen-7B2025.0957.642
BaseBackbone=DeepSeek-R1-Distill-Llama-8B2025.0963.249.3
BaseBackbone=DeepSeek-R1-Distill-Qwen-7B2025.0963.951.9