Helpfulness evaluation on HH-RLHF (test)
77Helpfulness FractionDeAL
Evaluation Results
| Method | Links | |
|---|---|---|
| DeALBackbone=MPT-7B-Instruct, Reward Model=R_helpful2024.02 | 77 | |
| DeALBackbone=MPT-7B-Instruct, Reward Model=R_hh2024.02 | 67 | |
| pa (for safety)Backbone=MPT-7B-Instruct2024.02 | 60 | |
| Helpful rerankBackbone=MPT-7B-Instruct2024.02 | 57 | |
| Harmless rerankBackbone=MPT-7B-Instruct2024.02 | 53 | |
| BaseBackbone=MPT-7B-Instruct2024.02 | 33 | |
| DeALBackbone=MPT-7B-Instruct, Reward Model=R_harmless2024.02 | 23 |