List your company

Safety Evaluation on WildChat

97.5Safe@1UnsafeChain
36.1452.076883.93Jul 29, 2025Sep 16, 2025Nov 4, 2025Dec 23, 2025Feb 10, 2026Mar 31, 2026May 20, 2026

Evaluation results

MethodLinks
UnsafeChainBase Model=R1-7B, Data Subset=full2025.0797.5
UnsafeChainBase Model=R1-8B, Data Subset=random2025.0796.5
UnsafeChainBase Model=R1-7B, Data Subset=random2025.0796
UnsafeChainBase Model=R1-7B, Data Subset=selected2025.0796
UnsafeChainBase Model=R1-1.5B, Data Subset=selected2025.0796
SafeChainBase Model=R1-8B2025.0795.5
STAR-1Base Model=R1-8B2025.0795.5
R1Model Size=7B2025.0795.5
STAR-1Base Model=R1-7B2025.0795.5
R1Model Size=8B2025.0795
UnsafeChainBase Model=R1-8B, Data Subset=selected2025.0795
SafeChainBase Model=R1-7B2025.0795
R1Model Size=1.5B2025.0795
SafeChainBase Model=R1-1.5B2025.0795
UnsafeChainBase Model=R1-1.5B, Data Subset=random2025.0795
UnsafeChainBase Model=R1-8B, Data Subset=full2025.0794.5
UnsafeChainBase Model=R1-1.5B, Data Subset=full2025.0793.5
STAR-1Base Model=R1-1.5B2025.0792
REFLECTOR (+GDPO)Backbone=Qwen-2.5-7B-Instruct2026.0587.8
REFLECTOR (SFT)Backbone=Qwen-2.5-7B-Instruct2026.0581.8
REFLECTOR (+GDPO)Backbone=Llama-3.1-8B-Instruct2026.0581.2
STAIRBackbone=Qwen-2.5-7B-Instruct2026.0577.8
REFLECTOR (+SFT)Backbone=Llama-3.1-8B-Instruct2026.0572.4
Shallow-AlignBackbone=Qwen-2.5-7B-Instruct2026.0571.9
STAIRBackbone=Llama-3.1-8B-Instruct2026.0569.86
Shallow-AlignBackbone=Llama-3.1-8B-Instruct2026.0564.2
DPOBackbone=Qwen-2.5-7B-Instruct2026.0553
Self-CritiqueBackbone=Llama-3.1-8B-Instruct2026.0547.5
Self-CritiqueBackbone=Qwen-2.5-7B-Instruct2026.0547.2
DPOBackbone=Llama-3.1-8B-Instruct2026.0544.79
SFTBackbone=Llama-3.1-8B-Instruct2026.0542.68
SFTBackbone=Qwen-2.5-7B-Instruct2026.0540.4
OriginalBackbone=Qwen-2.5-7B-Instruct2026.0539.6
OriginalBackbone=Llama-3.1-8B-Instruct2026.0538.5