Clinical error evaluation on ReEvalMed (test)
19D Error ScoreBLEU
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| BLEUType=NLP2026.06 | 19 | 19 | 19 | 0 | |
| BERTScoreType=NLP2026.06 | 24 | 24 | 24 | 0 | |
| RadBERTScoreType=Med.2026.06 | 26.5 | 26.5 | 26.5 | 0 | |
| RadGraphType=Med.2026.06 | 28.5 | 28.5 | 28.5 | 0 | |
| RaTEScoreType=Med.2026.06 | 35.5 | 35.5 | 35.5 | 0 | |
| CheXbertType=Med.2026.06 | 47 | 49 | 48 | 2 | |
| AlignScoreType=NLP2026.06 | 51 | 51 | 51 | 0 | |
| GREENType=LLM2026.06 | 53.5 | 41.5 | 47.5 | 12 | |
| CRIMSONType=LLM2026.06 | 75 | 80.5 | 77.8 | 5.5 | |
| RadFactType=LLM2026.06 | 80 | 56 | 68 | 24 | |
| FineRadScoreType=LLM2026.06 | 86.5 | 68.5 | 77.5 | 18 |