Instruction Following on MedMT-Text
55.79Instruction Following Score (ISR)ImpRIF-32B
Evaluation Results
| Method | Links | |
|---|---|---|
| ImpRIF-32BTraining Stage=SFT+RL2026.02 | 55.79 | |
| ImpRIF-32BTraining Stage=SFT2026.02 | 51.58 | |
| ImpRIF-8BTraining Stage=SFT+RL2026.02 | 48.07 | |
| ImpRIF-32BTraining Stage=RL2026.02 | 44.75 | |
| ImpRIF-8BTraining Stage=SFT2026.02 | 43.86 | |
| Qwen3-32BTraining Stage=Base2026.02 | 41.75 | |
| Qwen3-14BTraining Stage=Base2026.02 | 41.4 | |
| ImpRIF-8BTraining Stage=RL2026.02 | 40 | |
| ImpRIF-4BTraining Stage=SFT+RL2026.02 | 38.25 | |
| Qwen3-235B-A22BTraining Stage=Base2026.02 | 37.54 | |
| ImpRIF-4BTraining Stage=SFT2026.02 | 36.84 | |
| ImpRIF-4BTraining Stage=RL2026.02 | 36.84 | |
| Qwen3-8BTraining Stage=Base2026.02 | 34.39 | |
| Ministral3-14BTraining Stage=Base2026.02 | 33.68 | |
| Qwen2.5-72BReasoning Capability=Lacks inference reasoning capabilities2026.02 | 32.28 | |
| Qwen3-4BTraining Stage=Base2026.02 | 31.58 | |
| Llama3.1-70BReasoning Capability=Lacks inference reasoning capabilities2026.02 | 24.91 |