List your company

Prosocial Alignment on PKUSafeRLHF (test)

81.5MIPPROSOCIALALIGN
49.05257.47665.974.324Dec 6, 2025

Evaluation results

MethodLinks
PROSOCIALALIGNBackbone=llama2025.1281.5
PROATTR-GEN-PCABackbone=llama2025.1273.5
PPBackbone=llama2025.1271.7
PROSOCIALALIGNBackbone=mistral2025.1271.5
PV-ARM-SUMBackbone=llama2025.1269.2
PROATTR-GEN-PCABackbone=mistral2025.1268.7
DIREGBackbone=llama2025.1268.5
CTRL-GENBackbone=llama2025.1265.8
DIREGBackbone=mistral2025.1265.2
CTRL-GENBackbone=mistral2025.1264.3
PPBackbone=mistral2025.1262.7
SAFE-ARITHBackbone=llama2025.1256.6
SAFE-ARITHBackbone=mistral2025.1251.3
PV-ARM-SUMBackbone=mistral2025.1250.3