LLM Alignment on AlpacaEval IMDB 2.0 (test)
39.72LC WRCOALA
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| COALABackbone=Dolphin-7B2026.05 | 39.72 | 38.46 | 454 | |
| DPOBackbone=Dolphin-7B2026.05 | 33.86 | 31.58 | 511 | |
| COALABackbone=Llama-3.1-8B2026.05 | 27.64 | 25.68 | 415 | |
| ORPOBackbone=Dolphin-7B2026.05 | 24.9 | 22.92 | 448 | |
| COALABackbone=Mistral-7B2026.05 | 24.88 | 23.11 | 418 | |
| DPOBackbone=Mistral-7B2026.05 | 24.3 | 22.74 | 502 | |
| DPOBackbone=Llama-3.1-8B2026.05 | 21.79 | 20.18 | 449 | |
| ORPOBackbone=Mistral-7B2026.05 | 17.58 | 15.62 | 368 | |
| SFTBackbone=Dolphin-7B2026.05 | 16.21 | 15.47 | 423 | |
| ORPOBackbone=Llama-3.1-8B2026.05 | 12.1 | 12.05 | 610 | |
| SFTBackbone=Mistral-7B2026.05 | 8.42 | 1.77 | 428 | |
| SFTBackbone=Llama-3.1-8B2026.05 | 8.16 | 8.11 | 435 |