AlpacaEval 2.0 on UltraFeedback
30LCSSPO
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| SSPOBackbone=Mistral (7B), Size=10%2025.10 | 30 | 20.7 | |
| SSPOBackbone=Mistral (7B), Size=1%2025.10 | 26.7 | 18.1 | |
| SSPOBackbone=Llama3 (8B), Size=10%2025.10 | 20.7 | 20.8 | |
| SPABackbone=Mistral (7B), Size=10%2025.10 | 19.1 | 18.7 | |
| KTOBackbone=Mistral (7B), Size=10%2025.10 | 18.8 | 16.4 | |
| SPABackbone=Mistral (7B), Size=1%2025.10 | 18.2 | 15.6 | |
| SimPOBackbone=Mistral (7B), Size=10%2025.10 | 18.1 | 12.9 | |
| DPOBackbone=Mistral (7B), Size=10%2025.10 | 18 | 13.6 | |
| DPOBackbone=Mistral (7B), Size=1%2025.10 | 17 | 12.8 | |
| ORPOBackbone=Mistral (7B), Size=10%2025.10 | 16.7 | 10.9 | |
| KTOBackbone=Llama3 (8B), Size=10%2025.10 | 16.7 | 18.2 | |
| KTOBackbone=Mistral (7B), Size=1%2025.10 | 16.4 | 14.9 | |
| SSRMBackbone=Mistral (7B), Size=10%2025.10 | 16.2 | 13.3 | |
| SSRMBackbone=Llama3 (8B), Size=10%2025.10 | 15.1 | 16.3 | |
| ORPOBackbone=Mistral (7B), Size=1%2025.10 | 15 | 11.4 | |
| SSRMBackbone=Mistral (7B), Size=1%2025.10 | 14.9 | 13.2 | |
| SSRMBackbone=Llama3 (8B), Size=1%2025.10 | 14.9 | 12.8 | |
| SSPOBackbone=Llama3 (8B), Size=1%2025.10 | 14.8 | 16 | |
| SPABackbone=Llama3 (8B), Size=10%2025.10 | 14.5 | 16.6 | |
| KTOBackbone=Llama3 (8B), Size=1%2025.10 | 14.4 | 15.8 | |
| SimPOBackbone=Llama3 (8B), Size=1%2025.10 | 14.3 | 15.1 | |
| SPABackbone=Llama3 (8B), Size=1%2025.10 | 13.4 | 15.3 | |
| SimPOBackbone=Mistral (7B), Size=1%2025.10 | 13.2 | 8.3 | |
| DPOBackbone=Llama3 (8B), Size=10%2025.10 | 13 | 13.7 | |
| SimPOBackbone=Llama3 (8B), Size=10%2025.10 | 13 | 13.7 | |
| DPOBackbone=Llama3 (8B), Size=1%2025.10 | 12.1 | 12.6 | |
| ORPOBackbone=Llama3 (8B), Size=10%2025.10 | 10 | 10.9 | |
| ORPOBackbone=Llama3 (8B), Size=1%2025.10 | 9.4 | 10.3 | |
| SSPOBackbone=Phi-2 (2.7B), Size=10%2025.10 | 7.7 | 4.3 | |
| SSPOBackbone=Phi-2 (2.7B), Size=1%2025.10 | 7.2 | 4.1 | |
| SPABackbone=Phi-2 (2.7B), Size=10%2025.10 | 4.9 | 3.1 | |
| DPOBackbone=Phi-2 (2.7B), Size=10%2025.10 | 4.6 | 2.6 | |
| SSRMBackbone=Phi-2 (2.7B), Size=10%2025.10 | 4.4 | 2.6 | |
| KTOBackbone=Phi-2 (2.7B), Size=10%2025.10 | 4.4 | 2.6 | |
| SSRMBackbone=Phi-2 (2.7B), Size=1%2025.10 | 4.3 | 2 | |
| SimPOBackbone=Phi-2 (2.7B), Size=1%2025.10 | 4 | 2.5 | |
| SimPOBackbone=Phi-2 (2.7B), Size=10%2025.10 | 4 | 2.5 | |
| KTOBackbone=Phi-2 (2.7B), Size=1%2025.10 | 4 | 2.5 | |
| SPABackbone=Phi-2 (2.7B), Size=1%2025.10 | 4 | 2.6 | |
| ORPOBackbone=Phi-2 (2.7B), Size=10%2025.10 | 3.9 | 2.6 | |
| ORPOBackbone=Phi-2 (2.7B), Size=1%2025.10 | 3.7 | 2.3 | |
| DPOBackbone=Phi-2 (2.7B), Size=1%2025.10 | 3.6 | 2.5 |