Mathematical Reasoning on GSM8K (T, H, U, A)
71.9T ScoreTruthRL
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| TruthRLbase_model=Llama3.1-8B-Instruct2025.09 | 71.9 | 12.7 | 2.7 | 84.6 | |
| Promptingbase_model=Llama3.1-8B-Instruct2025.09 | 66.7 | 16.3 | 0.7 | 83 | |
| RFTbase_model=Llama3.1-8B-Instruct2025.09 | 59.8 | 19.6 | 0.9 | 79.5 | |
| R-Tuningbase_model=Llama3.1-8B-Instruct2025.09 | -4 | 6.4 | 91.3 | 2.4 | |
| SFTbase_model=Llama3.1-8B-Instruct2025.09 | -67.4 | 83.7 | 0 | 16.3 |