List your company

Safety Evaluation on Xs (test)

100Safety ScoreREFLECTOR (+SFT)
68.2876.51584.7592.985May 20, 2026

Evaluation results

MethodLinks
REFLECTOR (+SFT)Backbone=Llama-3.1-8B-Instruct2026.05100
REFLECTOR (+GDPO)Backbone=Llama-3.1-8B-Instruct2026.05100
REFLECTOR (SFT)Backbone=Qwen-2.5-7B-Instruct2026.05100
REFLECTOR (+GDPO)Backbone=Qwen-2.5-7B-Instruct2026.05100
STAIRBackbone=Llama-3.1-8B-Instruct2026.0599
STAIRBackbone=Qwen-2.5-7B-Instruct2026.0599
Shallow-AlignBackbone=Llama-3.1-8B-Instruct2026.0596.5
Shallow-AlignBackbone=Qwen-2.5-7B-Instruct2026.0596
SFTBackbone=Llama-3.1-8B-Instruct2026.0594.5
Self-CritiqueBackbone=Llama-3.1-8B-Instruct2026.0588.5
OriginalBackbone=Llama-3.1-8B-Instruct2026.0588
DPOBackbone=Llama-3.1-8B-Instruct2026.0586
SFTBackbone=Qwen-2.5-7B-Instruct2026.0584.5
Self-CritiqueBackbone=Qwen-2.5-7B-Instruct2026.0575.25
OriginalBackbone=Qwen-2.5-7B-Instruct2026.0573.5
DPOBackbone=Qwen-2.5-7B-Instruct2026.0569.5