Truthfulness on TruthfulQA (Reward, TR, IR, Div)
-1.8RewardSEA
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| SEAModel Backbone=LLaMA-3.2-Instruct (1B)2025.05 | -1.8 | 89 | 99 | 87 | |
| CBSModel Backbone=LLaMA-3.2-Base (3B)2025.05 | -2.59 | 67 | 100 | 87 | |
| SEAModel Backbone=LLaMA-3-Base (8B)2025.05 | -2.66 | 76 | 99 | 87 | |
| BoN-64Model Backbone=LLaMA-3.2-Base (3B)2025.05 | -2.73 | 74 | 100 | 85 | |
| SEAModel Backbone=LLaMA-3.2-Base (3B)2025.05 | -2.93 | 80 | 100 | 91 | |
| RSModel Backbone=LLaMA-3-Base (8B)2025.05 | -3 | 67 | 99 | 84 | |
| BoN-64Model Backbone=LLaMA-3.2-Instruct (1B)2025.05 | -3.01 | 77 | 98 | 83 | |
| BoN-64Model Backbone=LLaMA-3-Base (8B)2025.05 | -3.1 | 72 | 98 | 85 | |
| CBSModel Backbone=LLaMA-3.2-Instruct (1B)2025.05 | -3.17 | 75 | 97 | 85 | |
| CBSModel Backbone=LLaMA-3-Base (8B)2025.05 | -3.18 | 64 | 98 | 83 | |
| BoN-32Model Backbone=LLaMA-3.2-Base (3B)2025.05 | -3.25 | 73 | 98 | 87 | |
| RSModel Backbone=LLaMA-3.2-Base (3B)2025.05 | -3.37 | 66 | 100 | 86 | |
| BoN-32Model Backbone=LLaMA-3.2-Instruct (1B)2025.05 | -3.44 | 79 | 97 | 84 | |
| RSModel Backbone=LLaMA-3.2-Instruct (1B)2025.05 | -3.54 | 86 | 99 | 81 | |
| BoN-32Model Backbone=LLaMA-3-Base (8B)2025.05 | -3.55 | 68 | 97 | 85 | |
| SEAModel Backbone=LLaMA-3.2-Base (1B)2025.05 | -3.64 | 78 | 100 | 90 | |
| CBSModel Backbone=LLaMA-3.2-Base (1B)2025.05 | -3.95 | 67 | 100 | 86 | |
| BoN-64Model Backbone=LLaMA-3.2-Base (1B)2025.05 | -4.07 | 78 | 99 | 81 | |
| BoN-8Model Backbone=LLaMA-3.2-Base (3B)2025.05 | -4.08 | 76 | 99 | 87 | |
| SFTModel Backbone=LLaMA-3.2-Base (3B)2025.05 | -4.17 | 64 | 98 | 89 | |
| SFTModel Backbone=LLaMA-3-Base (8B)2025.05 | -4.23 | 62 | 100 | 86 | |
| BoN-32Model Backbone=LLaMA-3.2-Base (1B)2025.05 | -4.39 | 77 | 100 | 82 | |
| BoN-8Model Backbone=LLaMA-3.2-Instruct (1B)2025.05 | -4.39 | 76 | 95 | 83 | |
| BoN-8Model Backbone=LLaMA-3-Base (8B)2025.05 | -4.48 | 70 | 97 | 85 | |
| ARGSModel Backbone=LLaMA-3.2-Base (3B)2025.05 | -4.59 | 64 | 98 | 78 | |
| ARGSModel Backbone=LLaMA-3-Base (8B)2025.05 | -4.68 | 73 | 87 | 82 | |
| SFTModel Backbone=LLaMA-3.2-Instruct (1B)2025.05 | -4.87 | 72 | 95 | 85 | |
| ARGSModel Backbone=LLaMA-3.2-Instruct (1B)2025.05 | -4.87 | 72 | 99 | 53 | |
| RSModel Backbone=LLaMA-3.2-Base (1B)2025.05 | -5.26 | 66 | 98 | 83 | |
| ARGSModel Backbone=LLaMA-3.2-Base (1B)2025.05 | -5.31 | 55 | 97 | 76 | |
| BoN-8Model Backbone=LLaMA-3.2-Base (1B)2025.05 | -5.64 | 75 | 98 | 84 | |
| SFTModel Backbone=LLaMA-3.2-Base (1B)2025.05 | -6.14 | 59 | 98 | 86 |