Instruction Following Evaluation on IFEval Inverse
83.7AccuracyQwen3-30B
Evaluation results
| Method | Links | |
|---|---|---|
| Qwen3-30BTraining Variant=M-DPOP2025.10 | 83.7 | |
| Qwen3-4BTraining Variant=M-DPOP2025.10 | 76.1 | |
| GLM-4-9BTraining Variant=M-DPOP2025.10 | 74.2 | |
| GPT-5.2Training Variant=Zero-Shot2025.10 | 73.7 | |
| Gemini-3-FlashTraining Variant=Zero-Shot2025.10 | 70.6 | |
| Gemma-3-4BTraining Variant=M-DPOP2025.10 | 69.8 | |
| Claude-4.5-SonnetTraining Variant=Zero-Shot2025.10 | 67.2 | |
| Tulu-3.1-8BTraining Variant=M-DPOP2025.10 | 65.3 | |
| Qwen3-30BTraining Variant=GRPO2025.10 | 56.8 | |
| Qwen3-4BTraining Variant=GRPO2025.10 | 51.2 | |
| Qwen3-30BTraining Variant=Base2025.10 | 49.2 | |
| GLM-4-9BTraining Variant=GRPO2025.10 | 48.5 | |
| Qwen3-4BTraining Variant=Base2025.10 | 44.7 | |
| Gemma-3-4BTraining Variant=GRPO2025.10 | 43.7 | |
| GLM-4-9BTraining Variant=Base2025.10 | 42.1 | |
| Tulu-3.1-8BTraining Variant=GRPO2025.10 | 40.2 | |
| Gemma-3-4BTraining Variant=Base2025.10 | 38.5 | |
| Tulu-3.1-8BTraining Variant=Base2025.10 | 32.8 |