F1/Precision/Recall Evaluation on Medical Report Generation on MIMIC-CXR
70.8F1 ScoreTIF-GRPO
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| TIF-GRPOTraining Strategy=TIF-GRPO2026.05 | 70.8 | 74.7 | 67.5 | |
| DART2025.04 | 53.3 | 52 | 54.6 | |
| GRPO+LLMTraining Strategy=GRPO+LLM2026.05 | 52.7 | 60.9 | 46.7 | |
| I3 + C2FD2025.04 | 47.3 | 46.5 | 48.2 | |
| GRPO+ROUGETraining Strategy=GRPO+ROUGE2026.05 | 45.3 | 52.6 | 39.3 | |
| RGRG2025.04 | 44.7 | 46.1 | 47.5 | |
| MedGemmaParameters=4B, Reproduced=true2025.12 | 41.08 | 40.77 | 41.4 | |
| MRG-R12025.12 | 40.39 | 45.32 | 37.7 | |
| Med-LLM2025.04 | 39.5 | 41.2 | 37.3 | |
| DTrace2025.12 | 39.1 | 41.1 | 43.6 | |
| MA2025.04 | 38.9 | 41.1 | 39.8 | |
| RIHA2026.04 | 37.5 | 48.6 | 29.8 | |
| R2GenCMN2026.04 | 37.4 | 44 | 32.5 | |
| DCL2025.12 | 37.3 | 47.1 | 35.2 | |
| GSKET2025.12 | 37.1 | 45.8 | 34.8 | |
| MedGemmaParameters=27B, Reproduced=true2025.12 | 36.04 | 35.18 | 36.95 | |
| CXRMate2025.12 | 35.7 | 43.8 | 34.9 | |
| XProNet2026.04 | 35.3 | 46.3 | 28.5 | |
| PromptMRG2026.04 | 35.2 | 47.1 | 19.8 | |
| CheXagentReproduced=true2025.12 | 31.95 | 45.6 | 24.59 | |
| METransformer2025.04 | 31.1 | 36.4 | 30.9 | |
| BioMedGPTReproduced=true2025.12 | 28.6 | 29 | 31.4 | |
| SFTTraining Strategy=SFT2026.05 | 28.6 | 44.1 | 21.2 | |
| R2Gen2026.04 | 28 | 40.6 | 21.3 | |
| R2GenCMN2025.04 | 27.8 | 33.4 | 27.5 | |
| R2GenCMN2025.12 | 27.8 | 33.4 | 27.5 | |
| R2Gen2025.04 | 27.6 | 33.3 | 27.3 | |
| R2Gen2025.12 | 27.6 | 33.3 | 27.3 | |
| BaselineTraining Strategy=Baseline2026.05 | 25.3 | 21.1 | 33.1 | |
| ADDAATT2026.04 | 24.9 | 32.2 | 23.9 | |
| ATT2IN2026.04 | 20.4 | 26.8 | 20.3 | |
| HuatuoGPT-VisionReproduced=true2025.12 | 19.45 | 23.67 | 16.51 | |
| LLaVA-MedReproduced=true2025.12 | 16.64 | 26.99 | 12.03 | |
| RadFMReproduced=true2025.12 | 10.96 | 10.03 | 12.08 |