Open-ended Text Generation on CEB
80Sentiment ScoreFairSteer
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| FairSteerBackbone=Vicuna-13B2025.04 | 80 | 0.55 | 82 | 14.19 | |
| Llama3-8B2025.04 | 76 | 0.78 | 61 | 19.38 | |
| FairSteerBackbone=Llama2-13B2025.04 | 75 | 0.81 | 72 | 19.81 | |
| Vicuna-13B2025.04 | 72 | 1.31 | 69 | 20.31 | |
| FairSteerBackbone=Llama3-8B2025.04 | 70 | 0.71 | 65 | 19.22 | |
| FairSteerBackbone=Llama2-7B2025.04 | 69 | 1.44 | 62 | 20.42 | |
| Llama2-13B2025.04 | 69 | 1.05 | 51 | 23.44 | |
| Llama2-7B2025.04 | 68 | 1.19 | 54 | 21.24 | |
| FairSteerBackbone=Vicuna-7B2025.04 | 67 | 1.23 | 84 | 15.51 | |
| FairSteerBackbone=Mistral-7B2025.04 | 67 | 1.84 | 54 | 21.87 | |
| Vicuna-7B2025.04 | 66 | 1.78 | 69 | 17.74 | |
| Mistral-7B2025.04 | 66 | 1.48 | 53 | 22.84 |