General Reasoning and Creative Writing on ArenaHard v2
15.5Hard Prompt ScoreSDPO
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| SDPOBase Model=Qwen3-8B2026.02 | 15.5 | 16.2 | |
| Qwen3-8BModel Size=8B2026.02 | 14 | 13.7 | |
| Qwen3-4BModel Size=4B2026.02 | 9 | 8 | |
| SDPOBase Model=Qwen3-4B2026.02 | 7.8 | 7.9 | |
| Olmo3-7B-SFTModel Size=7B, Training=SFT2026.02 | 2.4 | 1.4 | |
| SDPOBase Model=Olmo3-7B-SFT2026.02 | 2.4 | 1.4 | |
| SDPOBase Model=Olmo3-7B-DPO2026.02 | 2 | 10 | |
| Olmo3-7B-DPOModel Size=7B, Training=DPO2026.02 | 1.7 | 8.2 |