LLM Alignment on AlpacaEval ULTRA 2.0 (test)
31.58LC WR (%)COALA
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| COALABackbone=Dolphin-7B2026.05 | 31.58 | 30.21 | 445 | |
| DPOBackbone=Dolphin-7B2026.05 | 26.41 | 24.86 | 476 | |
| ORPOBackbone=Dolphin-7B2026.05 | 22.94 | 25.93 | 452 | |
| COALABackbone=Mistral-7B2026.05 | 20.84 | 20.91 | 459 | |
| COALABackbone=Llama-3.1-8B2026.05 | 20.64 | 18.32 | 552 | |
| DPOBackbone=Llama-3.1-8B2026.05 | 18.89 | 15.81 | 503 | |
| DPOBackbone=Mistral-7B2026.05 | 17.68 | 15.56 | 453 | |
| ORPOBackbone=Mistral-7B2026.05 | 16.04 | 12.28 | 350 | |
| SFTBackbone=Dolphin-7B2026.05 | 14.88 | 12.76 | 459 | |
| ORPOBackbone=Llama-3.1-8B2026.05 | 12.91 | 10.95 | 354 | |
| SFTBackbone=Llama-3.1-8B2026.05 | 7.41 | 5.62 | 546 | |
| SFTBackbone=Mistral-7B2026.05 | 6.18 | 6.11 | 463 |