List your company

Safety alignment evaluation on Llama-Guard

82.14Harmfulness (%)IPO
81.438886.171990.90595.6381Nov 28, 2025

Evaluation results

MethodLinks
IPOBackbone=Llama3.1-8B-Base2025.1182.14
DPOBackbone=Llama3.1-8B-Base2025.1183.06
KTOBackbone=Llama3.1-8B-Base2025.1183.42
DPO-RandomBackbone=Llama3.1-8B-Base2025.1183.42
SimPOBackbone=Llama3.1-8B-Base2025.1184.15
RTOBackbone=Llama3.1-8B-Base2025.1185.71
TDPOBackbone=Llama3.1-8B-Base2025.1186.54
TIS-DPOBackbone=Llama3.1-8B-Base2025.1187.87
AAOBackbone=Llama3.1-8B-Base2025.1191.29
RTOBackbone=Mistral-7B-Base2025.1194.36
TDPOBackbone=Mistral-7B-Base2025.1194.55
KTOBackbone=Mistral-7B-Base2025.1195.04
DPOBackbone=Mistral-7B-Base2025.1195.48
DPO-RandomBackbone=Mistral-7B-Base2025.1195.52
DPOBackbone=Llama3.1-8B-Instruct2025.1196.12
IPOBackbone=Mistral-7B-Base2025.1196.34
KTOBackbone=Llama3.1-8B-Instruct2025.1196.52
SimPOBackbone=Mistral-7B-Base2025.1196.56
TIS-DPOBackbone=Mistral-7B-Base2025.1196.74
RTOBackbone=Llama3.1-8B-Instruct2025.1196.99
TDPOBackbone=Llama3.1-8B-Instruct2025.1197.04
DPO-RandomBackbone=Llama3.1-8B-Instruct2025.1197.32
TIS-DPOBackbone=Mistral-7B-Instruct2025.1197.35
AAOBackbone=Mistral-7B-Base2025.1197.41
IPOBackbone=Llama3.1-8B-Instruct2025.1197.42
TIS-DPOBackbone=Llama3.1-8B-Instruct2025.1197.54
SimPOBackbone=Llama3.1-8B-Instruct2025.1197.85
TDPOBackbone=Mistral-7B-Instruct2025.1197.97
AAOBackbone=Llama3.1-8B-Instruct2025.1198.42
DPO-RandomBackbone=Mistral-7B-Instruct2025.1198.42
RTOBackbone=Mistral-7B-Instruct2025.1198.76
DPOBackbone=Mistral-7B-Instruct2025.1199.14
KTOBackbone=Mistral-7B-Instruct2025.1199.21
SimPOBackbone=Mistral-7B-Instruct2025.1199.35
IPOBackbone=Mistral-7B-Instruct2025.1199.42
AAOBackbone=Mistral-7B-Instruct2025.1199.67