Question Answering on TruthfulQA (BLEU/ROUGE/BLEURT Metrics)
0.748BLEURTThree-Stage Fine-Tuning Method
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| Three-Stage Fine-Tuning MethodInstruction Model=Qwen2.5-14B2025.04 | 0.748 | 0.615 | 0.642 | |
| SFTInstruction Model=Qwen2.5-14B2025.04 | 0.721 | 0.577 | 0.613 | |
| OriginalInstruction Model=Qwen2.5-14B2025.04 | 0.709 | 0.582 | 0.609 | |
| Three-Stage Fine-Tuning MethodInstruction Model=Qwen2-7B2025.04 | 0.678 | 0.551 | 0.589 | |
| Three-Stage Fine-Tuning MethodInstruction Model=Llama3-8B2025.04 | 0.668 | 0.528 | 0.586 | |
| OriginalInstruction Model=Qwen2-7B2025.04 | 0.636 | 0.543 | 0.57 | |
| SFTInstruction Model=Qwen2-7B2025.04 | 0.633 | 0.562 | 0.577 | |
| Three-Stage Fine-Tuning MethodInstruction Model=Llama3.2-3B2025.04 | 0.632 | 0.534 | 0.556 | |
| OriginalInstruction Model=Llama3-8B2025.04 | 0.625 | 0.489 | 0.553 | |
| SFTInstruction Model=Llama3.2-3B2025.04 | 0.617 | 0.523 | 0.521 | |
| OriginalInstruction Model=Llama3.2-3B2025.04 | 0.613 | 0.517 | 0.527 | |
| SFTInstruction Model=Llama3-8B2025.04 | 0.607 | 0.503 | 0.562 |