Truthfulness Evaluation on TruthfulQA (avg.@8)
68.69Average Score (@8)Llama3.1-8B-Instruct
Evaluation Results
| Method | Links | |
|---|---|---|
| Llama3.1-8B-InstructTraining Pipeline=PSFT2025.08 | 68.69 | |
| Qwen2.5-7B-InstructTraining Pipeline=PSFT2025.08 | 67.16 | |
| Llama3.1-8B-InstructTraining Pipeline=SFT2025.08 | 67.08 | |
| Llama3.1-8B-InstructTraining Pipeline=PSFT → GRPO2025.08 | 65.24 | |
| Qwen2.5-7B-InstructTraining Pipeline=PSFT → GRPO2025.08 | 64.84 | |
| Llama3.1-8B-InstructTraining Pipeline=SFT → GRPO2025.08 | 64.82 | |
| Qwen2.5-7B-InstructTraining Pipeline=SFT2025.08 | 63.14 | |
| Qwen2.5-7B-InstructTraining Pipeline=SFT → GRPO2025.08 | 61.93 |