Instruction Following on HaluEval (test)
0.36Failure Rate (Sum)Three-Stage Fine-Tuning Method
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| Three-Stage Fine-Tuning MethodInstruction Model=Llama3-8B2025.04 | 0.36 | 0.43 | 0.69 | |
| SFTInstruction Model=Llama3-8B2025.04 | 2.17 | 0.58 | 0.95 | |
| OriginalInstruction Model=Llama3-8B2025.04 | 2.61 | 0.63 | 1.03 | |
| Three-Stage Fine-Tuning MethodInstruction Model=Llama3.2-3B2025.04 | 3.55 | 1.24 | 1.09 | |
| SFTInstruction Model=Llama3.2-3B2025.04 | 8.73 | 2.17 | 2.36 | |
| OriginalInstruction Model=Llama3.2-3B2025.04 | 13.67 | 1.96 | 3.54 |