Instruction Following on WildBench (test)
58.6Info SeekGPT-4o
Evaluation Results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| GPT-4oModel Type=Proprietary2026.01 | 58.6 | 59.1 | 60.5 | 60.2 | 57.3 | 59.3 | |
| GPT-4-turboModel Type=Proprietary2026.01 | 57.2 | 58.7 | 55.1 | 56.2 | 51 | 55.2 | |
| DeepSeek-V2-ChatModel Type=Chat2026.01 | 52.7 | 56.4 | 55 | 54.8 | 51.4 | 54 | |
| OriginalBackbone=Qwen-2.5-7B, Setting=Instruct2026.01 | 47.2 | 44.5 | 44.7 | 48.9 | 44.4 | 45.4 | |
| STEPSBackbone=Llama-3-8B, Setting=Instruct2026.01 | 46.1 | 25.2 | 37.9 | 42.9 | 18.9 | 36.3 | |
| STEPSBackbone=Qwen-2.5-7B, Setting=Instruct2026.01 | 44 | 45.1 | 49.1 | 48.3 | 45.5 | 45.7 | |
| Instruct-SkillMixBackbone=Qwen-2.5-7B, Setting=Instruct2026.01 | 41.9 | 36.6 | 41.7 | 40.8 | 37 | 39.6 | |
| STEPSBackbone=Mistral-7B-v0.3, Setting=Instruct2026.01 | 39.6 | 16.5 | 30.8 | 37.8 | 10 | 29 | |
| OriginalBackbone=Mistral-7B-v0.3, Setting=Instruct2026.01 | 39.5 | 20.7 | 29.3 | 33.7 | 8.5 | 28.6 | |
| STEPSBackbone=Qwen-2.5-7B, Setting=Base2026.01 | 39.3 | 34.5 | 31.1 | 39.5 | 34 | 35.1 | |
| Instruct-SkillMixBackbone=Llama-3-8B, Setting=Instruct2026.01 | 39.1 | 20.6 | 32.5 | 37.7 | 15.7 | 30.8 | |
| OriginalBackbone=Llama-3-8B, Setting=Instruct2026.01 | 37.9 | 43 | 26 | 35.8 | 22.4 | 34.2 | |
| Instruct-SkillMixBackbone=Qwen-2.5-7B, Setting=Base2026.01 | 35.3 | 33.8 | 27.9 | 36.1 | 31.7 | 32.8 | |
| STEPSBackbone=Llama-3-8B, Setting=Base2026.01 | 30.6 | 7.7 | 26.8 | 30.6 | 12.3 | 23 | |
| Instruct-SkillMixBackbone=Llama-3-8B, Setting=Base2026.01 | 22.4 | 13.8 | 23.9 | 26.4 | 9.7 | 19.7 | |
| STEPSBackbone=Mistral-7B-v0.3, Setting=Base2026.01 | 21.9 | -6.7 | 17.1 | 24.3 | -5.2 | 11.3 | |
| OriginalBackbone=Qwen-2.5-7B, Setting=Base2026.01 | 21.5 | 6.5 | 5.3 | 12.5 | 5.5 | 10.7 | |
| Instruct-SkillMixBackbone=Mistral-7B-v0.3, Setting=Instruct2026.01 | 18.6 | 2.3 | 12.1 | 11.8 | -3.1 | 10.3 | |
| Alpaca52kBackbone=Llama-3-8B, Setting=Instruct2026.01 | 15.9 | -10.4 | 4.1 | 11.3 | -9.1 | 5 | |
| Instruct-SkillMixBackbone=Mistral-7B-v0.3, Setting=Base2026.01 | 9.3 | -6.5 | 9.3 | 14 | -7.8 | 5.2 | |
| Alpaca52kBackbone=Llama-3-8B, Setting=Base2026.01 | 0.5 | -18.6 | -6.5 | 0.4 | -16.6 | -6.1 | |
| Alpaca52kBackbone=Qwen-2.5-7B, Setting=Base2026.01 | -6.3 | -5.4 | -2 | 3.3 | -7.7 | -3.8 | |
| Alpaca52kBackbone=Qwen-2.5-7B, Setting=Instruct2026.01 | -13.4 | -14.3 | -9.1 | -6.5 | -14.1 | -11.4 | |
| Alpaca52kBackbone=Mistral-7B-v0.3, Setting=Instruct2026.01 | -13.6 | -26.6 | -13.2 | -8.2 | -12.5 | -14.8 | |
| Alpaca52kBackbone=Mistral-7B-v0.3, Setting=Base2026.01 | -14.2 | -24.8 | -13.7 | -5.1 | -20 | -14 | |
| OriginalBackbone=Mistral-7B-v0.3, Setting=Base2026.01 | -26.1 | -28.1 | -21.2 | -9.3 | -29.8 | -22.7 | |
| OriginalBackbone=Llama-3-8B, Setting=Base2026.01 | -27.7 | -30.4 | -23.8 | -17.1 | -29.8 | -25.7 |