Prosocial Alignment on NicheHazardQA (test)
78.2MIPPROSOCIALALIGN
Evaluation Results
| Method | Links | |
|---|---|---|
| PROSOCIALALIGNBackbone=llama2025.12 | 78.2 | |
| PPBackbone=llama2025.12 | 70.1 | |
| PROATTR-GEN-PCABackbone=mistral2025.12 | 68.8 | |
| PROSOCIALALIGNBackbone=mistral2025.12 | 68.1 | |
| DIREGBackbone=llama2025.12 | 67.4 | |
| PV-ARM-SUMBackbone=llama2025.12 | 64.1 | |
| PROATTR-GEN-PCABackbone=llama2025.12 | 63.9 | |
| DIREGBackbone=mistral2025.12 | 59.4 | |
| PPBackbone=mistral2025.12 | 58.6 | |
| SAFE-ARITHBackbone=llama2025.12 | 53.9 | |
| CTRL-GENBackbone=llama2025.12 | 51.8 | |
| CTRL-GENBackbone=mistral2025.12 | 51.8 | |
| SAFE-ARITHBackbone=mistral2025.12 | 45.3 | |
| PV-ARM-SUMBackbone=mistral2025.12 | 44.2 |