Instruction Following on IFEval Inverse
83.7Inv-IFMulti-DPOP
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| Multi-DPOPModel=Qwen3-30B2025.10 | 83.7 | — | |
| Gemini-3-FlashTier=High2026.06 | 80.9 | — | |
| Gemini-3-pro High2026.06 | 79.6 | — | |
| Seed2.0 Pro2026.06 | 78.9 | — | |
| Seed2.0 Lite2026.06 | 77.1 | — | |
| Multi-DPOPModel=Qwen3-4B2025.10 | 76.1 | — | |
| GPT-5-miniTier=High2026.06 | 74.8 | — | |
| Multi-DPOPModel=GLM-4-9B2025.10 | 74.2 | — | |
| Claude-Opus-4.52026.06 | 72.4 | — | |
| GPT-5.2 High2026.06 | 72.3 | — | |
| Multi-DPOPModel=Gemma-3-4B2025.10 | 69.8 | — | |
| Seed2.0 Mini2026.06 | 69.3 | — | |
| Claude-Sonnet-4.52026.06 | 69.3 | — | |
| Multi-DPOPModel=Tulu-3.1-8B2025.10 | 65.3 | — | |
| GRPOModel=Qwen3-30B2025.10 | 56.8 | — | |
| DPOModel=Qwen3-30B2025.10 | 55.1 | — | |
| SFTModel=Qwen3-30B2025.10 | 51.4 | — | |
| GRPOModel=Qwen3-4B2025.10 | 51.2 | — | |
| DPOModel=Qwen3-4B2025.10 | 49.8 | — | |
| BaseModel=Qwen3-30B2025.10 | 49.2 | — | |
| GRPOModel=GLM-4-9B2025.10 | 48.5 | — | |
| DPOModel=GLM-4-9B2025.10 | 47.3 | — | |
| SFTModel=Qwen3-4B2025.10 | 46.3 | — | |
| SFTModel=GLM-4-9B2025.10 | 44.8 | — | |
| BaseModel=Qwen3-4B2025.10 | 44.7 | — | |
| GRPOModel=Gemma-3-4B2025.10 | 43.7 | — | |
| DPOModel=Gemma-3-4B2025.10 | 42.1 | — | |
| BaseModel=GLM-4-9B2025.10 | 42.1 | — | |
| GRPOModel=Tulu-3.1-8B2025.10 | 40.2 | — | |
| SFTModel=Gemma-3-4B2025.10 | 40.2 | — | |
| BaseModel=Gemma-3-4B2025.10 | 38.5 | — | |
| DPOModel=Tulu-3.1-8B2025.10 | 38.4 | — | |
| SFTModel=Tulu-3.1-8B2025.10 | 35.1 | — | |
| BaseModel=Tulu-3.1-8B2025.10 | 32.8 | — | |
| A-DPOBackbone=Tulu-3.1-8B2025.10 | — | 62.7 | |
| A-DPOBackbone=Gemma-3-4B2025.10 | — | 67.4 | |
| A-DPOBackbone=GLM-4-9B2025.10 | — | 71.8 | |
| A-DPOBackbone=Qwen3-4B2025.10 | — | 73.6 | |
| A-DPOBackbone=Qwen3-30B2025.10 | — | 80.5 | |
| Base ModelBackbone=Tulu-3.1-8B2025.10 | — | 32.8 | |
| Base ModelBackbone=Gemma-3-4B2025.10 | — | 38.5 | |
| Base ModelBackbone=GLM-4-9B2025.10 | — | 42.1 | |
| Base ModelBackbone=Qwen3-4B2025.10 | — | 44.7 | |
| Base ModelBackbone=Qwen3-30B2025.10 | — | 49.2 | |
| Claude-4-OpusSetting=Zero-Shot2025.10 | — | 67.2 | |
| DPOBackbone=Tulu-3.1-8B2025.10 | — | 38.4 | |
| DPOBackbone=Gemma-3-4B2025.10 | — | 42.1 | |
| DPOBackbone=GLM-4-9B2025.10 | — | 47.3 | |
| DPOBackbone=Qwen3-4B2025.10 | — | 49.8 | |
| DPOBackbone=Qwen3-30B2025.10 | — | 55.1 | |
| Gemini-2.5-ProSetting=Zero-Shot2025.10 | — | 70.6 | |
| GPT-5-highSetting=Zero-Shot2025.10 | — | 73.7 | |
| GRPOBackbone=Tulu-3.1-8B2025.10 | — | 40.2 | |
| GRPOBackbone=Gemma-3-4B2025.10 | — | 43.7 | |
| GRPOBackbone=GLM-4-9B2025.10 | — | 48.5 | |
| GRPOBackbone=Qwen3-4B2025.10 | — | 51.2 | |
| GRPOBackbone=Qwen3-30B2025.10 | — | 56.8 | |
| Multi-DPOPBackbone=Tulu-3.1-8B2025.10 | — | 65.3 | |
| Multi-DPOPBackbone=Gemma-3-4B2025.10 | — | 69.8 | |
| Multi-DPOPBackbone=GLM-4-9B2025.10 | — | 74.2 | |
| Multi-DPOPBackbone=Qwen3-4B2025.10 | — | 76.1 | |
| Multi-DPOPBackbone=Qwen3-30B2025.10 | — | 83.7 | |
| SFTBackbone=Tulu-3.1-8B2025.10 | — | 35.1 | |
| SFTBackbone=Gemma-3-4B2025.10 | — | 40.2 | |
| SFTBackbone=GLM-4-9B2025.10 | — | 44.8 | |
| SFTBackbone=Qwen3-4B2025.10 | — | 46.3 | |
| SFTBackbone=Qwen3-30B2025.10 | — | 51.4 |