Natural Language Inference on ANLI and HANS (unseen)
32.4ANLI ScoreQwen 2.5 7B
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| Qwen 2.5 7BK=7, type=Baseline2025.05 | 32.4 | 84.4 | 58.4 | |
| FOVER-Qwen2.5-7B-PRMK=7, Base Model=Qwen 2.5 7B2025.05 | 30.8 | 86 | 58.4 | |
| FOVER-Llama3.1-8B-PRMK=7, Base Model=Llama 3.1 8B2025.05 | 30 | 81.2 | 55.6 | |
| Qwen2.5-Math-7B-PRM800KK=7, Base Model=Qwen 2.5 7B2025.05 | 29.6 | 83.2 | 56.4 | |
| Qwen2.5-7B-Skywork-PRMK=7, Base Model=Qwen 2.5 7B2025.05 | 29.2 | 81.2 | 55.2 | |
| Qwen2.5-Math-PRM-7BK=7, Base Model=Qwen 2.5 7B2025.05 | 27.6 | 86 | 56.8 | |
| RLHFlow-Llama3.1-8B-DK=7, Base Model=Llama 3.1 8B2025.05 | 27.2 | 74.4 | 50.8 | |
| RLHFlow-Llama3.1-8B-MK=7, Base Model=Llama 3.1 8B2025.05 | 26.8 | 77.6 | 52.2 | |
| Llama 3.1 8BK=7, type=Baseline2025.05 | 26.4 | 69.2 | 47.8 |