Constraint-following Instruction Evaluation on IFEval
54.4Average ScoreLLAMA3-INSTRUCT w/ ULTRAFEEDBACK
Evaluation results
| Method | Links | |
|---|---|---|
| LLAMA3-INSTRUCT w/ ULTRAFEEDBACKBackbone=LLAMA3-INSTRUCT2024.10 | 54.4 | |
| LLAMA3-INSTRUCT w/ SSO_DPOBackbone=LLAMA3-INSTRUCT2024.10 | 53.4 | |
| LLAMA3-INSTRUCT w/ PBAA_DPOBackbone=LLAMA3-INSTRUCT2024.10 | 53.2 | |
| LLAMA3-INSTRUCTBackbone=LLAMA3-INSTRUCT2024.10 | 53 | |
| QWEN2-INSTRUCT w/ ULTRAFEEDBACKBackbone=QWEN2-INSTRUCT2024.10 | 51.5 | |
| QWEN2-INSTRUCTBackbone=QWEN2-INSTRUCT2024.10 | 51.4 | |
| QWEN2-INSTRUCT w/ SSO_DPOBackbone=QWEN2-INSTRUCT2024.10 | 51.4 | |
| QWEN2-INSTRUCT w/ PBAA_DPOBackbone=QWEN2-INSTRUCT2024.10 | 50.9 | |
| LLAMA3-SFT w/ SSO_DPOBackbone=LLAMA3-SFT2024.10 | 50.3 | |
| LLAMA3-SFT w/ PBAA_DPOBackbone=LLAMA3-SFT2024.10 | 47.8 | |
| QWEN2-SFT w/ SSO_DPOBackbone=QWEN2-SFT2024.10 | 45.7 | |
| LLAMA3-SFT w/ ULTRAFEEDBACKBackbone=LLAMA3-SFT2024.10 | 43.6 | |
| QWEN2-SFT w/ PBAA_DPOBackbone=QWEN2-SFT2024.10 | 43.6 | |
| QWEN2-SFT w/ ULTRAFEEDBACKBackbone=QWEN2-SFT2024.10 | 40.4 | |
| LLAMA3-SFTBackbone=LLAMA3-SFT2024.10 | 24.9 | |
| QWEN2-SFTBackbone=QWEN2-SFT2024.10 | 19.8 |