MT-Bench on UltraMedical Preference
6.9MT ScoreSSPO
Evaluation Results
| Method | Links | |
|---|---|---|
| SSPOBackbone=Llama3 (8B), Size=10%2025.10 | 6.9 | |
| SSPOBackbone=Llama3 (8B), Size=1%2025.10 | 6.7 | |
| DPOBackbone=Llama3 (8B), Size=1%2025.10 | 6.5 | |
| SimPOBackbone=Llama3 (8B), Size=10%2025.10 | 6.5 | |
| DPOBackbone=Llama3 (8B), Size=10%2025.10 | 6.4 | |
| ORPOBackbone=Llama3 (8B), Size=1%2025.10 | 6.4 | |
| ORPOBackbone=Llama3 (8B), Size=10%2025.10 | 6.4 | |
| SimPOBackbone=Llama3 (8B), Size=1%2025.10 | 6.4 | |
| SSRMBackbone=Llama3 (8B), Size=1%2025.10 | 6.4 | |
| SSRMBackbone=Llama3 (8B), Size=10%2025.10 | 6.4 | |
| KTOBackbone=Llama3 (8B), Size=1%2025.10 | 6.4 | |
| KTOBackbone=Llama3 (8B), Size=10%2025.10 | 6.4 | |
| SPABackbone=Llama3 (8B), Size=1%2025.10 | 6.4 | |
| SPABackbone=Llama3 (8B), Size=10%2025.10 | 6.4 | |
| DPOBackbone=Mistral (7B), Size=10%2025.10 | 5.3 | |
| SimPOBackbone=Mistral (7B), Size=1%2025.10 | 5.3 | |
| SimPOBackbone=Mistral (7B), Size=10%2025.10 | 5.3 | |
| SSRMBackbone=Mistral (7B), Size=10%2025.10 | 5.3 | |
| KTOBackbone=Mistral (7B), Size=10%2025.10 | 5.3 | |
| SPABackbone=Mistral (7B), Size=10%2025.10 | 5.3 | |
| SSPOBackbone=Mistral (7B), Size=10%2025.10 | 5.3 | |
| DPOBackbone=Mistral (7B), Size=1%2025.10 | 5.2 | |
| ORPOBackbone=Mistral (7B), Size=1%2025.10 | 5.2 | |
| ORPOBackbone=Mistral (7B), Size=10%2025.10 | 5.2 | |
| SSRMBackbone=Mistral (7B), Size=1%2025.10 | 5.2 | |
| KTOBackbone=Mistral (7B), Size=1%2025.10 | 5.2 | |
| SPABackbone=Mistral (7B), Size=1%2025.10 | 5.2 | |
| SSPOBackbone=Mistral (7B), Size=1%2025.10 | 5.2 |