Prosocial Alignment on PKUSafeRLHF (test)
81.5MIPPROSOCIALALIGN
Evaluation Results
| Method | Links | |
|---|---|---|
| PROSOCIALALIGNBackbone=llama2025.12 | 81.5 | |
| PROATTR-GEN-PCABackbone=llama2025.12 | 73.5 | |
| PPBackbone=llama2025.12 | 71.7 | |
| PROSOCIALALIGNBackbone=mistral2025.12 | 71.5 | |
| PV-ARM-SUMBackbone=llama2025.12 | 69.2 | |
| PROATTR-GEN-PCABackbone=mistral2025.12 | 68.7 | |
| DIREGBackbone=llama2025.12 | 68.5 | |
| CTRL-GENBackbone=llama2025.12 | 65.8 | |
| DIREGBackbone=mistral2025.12 | 65.2 | |
| CTRL-GENBackbone=mistral2025.12 | 64.3 | |
| PPBackbone=mistral2025.12 | 62.7 | |
| SAFE-ARITHBackbone=llama2025.12 | 56.6 | |
| SAFE-ARITHBackbone=mistral2025.12 | 51.3 | |
| PV-ARM-SUMBackbone=mistral2025.12 | 50.3 |