General Language Capabilities on Open LLM Leaderboard (test)
83.29HellaSwag AccuracyTDPO
Evaluation results
| Method | Links | |||||||
|---|---|---|---|---|---|---|---|---|
| TDPOBackbone=Llama3 8B2026.05 | 83.29 | 59.04 | 61.86 | 51.76 | 76.55 | 75.1 | 67.91 | |
| TIS-DPOBackbone=Mistral 7B v0.12026.05 | 83.1 | 59.09 | 57.81 | 45.91 | 77.66 | 34.42 | 59.66 | |
| TBPO-ABackbone=Mistral 7B v0.12026.05 | 82.74 | 59.47 | 60.54 | 44.4 | 78.21 | 39.04 | 60.73 | |
| TBPO-QBackbone=Mistral 7B v0.12026.05 | 82.72 | 59.74 | 60.82 | 44.47 | 78.58 | 39.34 | 60.95 | |
| TDPOBackbone=Mistral 7B v0.12026.05 | 82.7 | 61.54 | 57.17 | 43.48 | 76.95 | 28.65 | 58.41 | |
| DPOBackbone=Mistral 7B v0.12026.05 | 82.39 | 59.4 | 59.15 | 42.3 | 77.74 | 34.87 | 59.3 | |
| TBPO-ABackbone=Llama3 8B2026.05 | 81.97 | 64.08 | 64.84 | 53.29 | 79.63 | 79.22 | 70.5 | |
| TBPO-QBackbone=Llama3 8B2026.05 | 81.9 | 64.07 | 64.92 | 53.36 | 79.08 | 78.92 | 70.38 | |
| BPOBackbone=Llama3 8B2026.05 | 81.68 | 54.95 | 61.48 | 47.88 | 75.13 | 73.22 | 65.72 | |
| TIS-DPOBackbone=Llama3 8B2026.05 | 81.37 | 58.7 | 60.77 | 49.87 | 75.14 | 75.12 | 66.82 | |
| DPOBackbone=Llama3 8B2026.05 | 81.23 | 56.05 | 61.7 | 48.47 | 75.84 | 74.2 | 66.24 | |
| BPOBackbone=Mistral 7B v0.12026.05 | 81.03 | 57.67 | 59.16 | 41.71 | 77.26 | 32.9 | 58.29 | |
| SFTBackbone=Mistral 7B v0.12026.05 | 80.72 | 55.54 | 58.41 | 43.67 | 76.71 | 18.49 | 55.59 | |
| SFTBackbone=Llama3 8B2026.05 | 79.46 | 51.45 | 61.19 | 46.46 | 75.69 | 68.67 | 63.82 |