Value Alignment on Ethics (test)
5.57Align ScoreAEM + VM
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| AEM + VMFramework=SECOND THOUGHTS2023.01 | 5.57 | 34.73 | 5.57 | 6.29 | |
| AEM + AILFramework=SECOND THOUGHTS2023.01 | 5.04 | 32.09 | 6.22 | 5.38 | |
| DExperts2023.01 | 4.3 | 30.37 | 5.38 | 8.6 | |
| PPLM2023.01 | 3.97 | 26.53 | 5.58 | 5.25 | |
| Data Filtering2023.01 | 3.9 | 23.6 | 5.58 | 5.1 | |
| AEM OnlyFramework=SECOND THOUGHTS2023.01 | 3.86 | 31.41 | 5.12 | 5.75 | |
| InstructGPTSource=Huge LM API service2023.01 | 3.06 | 20.18 | 5.38 | 8.04 | |
| GPT-3Source=Huge LM API service2023.01 | 2.96 | 19.22 | 5.31 | 7.49 | |
| Safe Beam Search2023.01 | 2.66 | 19.82 | 5.08 | 10.31 | |
| MLE2023.01 | 2.11 | 17.32 | 5.57 | 5.23 |