LLM Alignment on AlpacaEval EDU 2.0 (test)
40.9LC WRCOALA
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| COALABackbone=Llama-3.1-8B2026.05 | 40.9 | 38.2 | 562 | |
| COALABackbone=Dolphin-7B2026.05 | 40.81 | 39.05 | 439 | |
| DPOBackbone=Llama-3.1-8B2026.05 | 40.68 | 38.53 | 539 | |
| DPOBackbone=Dolphin-7B2026.05 | 34.73 | 32.46 | 494 | |
| ORPOBackbone=Dolphin-7B2026.05 | 25.06 | 23.59 | 526 | |
| COALABackbone=Mistral-7B2026.05 | 24.61 | 23.82 | 592 | |
| DPOBackbone=Mistral-7B2026.05 | 24.19 | 22.45 | 492 | |
| ORPOBackbone=Llama-3.1-8B2026.05 | 23.87 | 20.58 | 599 | |
| SFTBackbone=Dolphin-7B2026.05 | 17.36 | 15.3 | 404 | |
| ORPOBackbone=Mistral-7B2026.05 | 17.01 | 14.67 | 561 | |
| SFTBackbone=Llama-3.1-8B2026.05 | 10.92 | 10.75 | 384 | |
| SFTBackbone=Mistral-7B2026.05 | 6.8 | 10.3 | 515 |