Named Entity Recognition on MIMIC-III (test) benchmark leaderboard · SOTA2 Research

Named Entity Recognition on MIMIC-III (test)

85.6Strict F1Centralized fine-tuning

Evaluation Results

Method
Centralized fine-tuning2026.01	85.6	95.4
Centralized fine-tuning2026.01	85.2	95.2
Fed-MedLoRA+2026.01	85	94.9
Fed-MedLoRA2026.01	84.7	94.2
Fed-MedLoRA+2026.01	84.5	94.5
Fed-MedLoRA2026.01	84.1	93.2
LLaMA3-8B Centralized2026.01	84	92.5
DeepSeek-R1-Distill-8B Centralized2026.01	83	92.5
Bio_ClinicalBERT Centralized2026.01	81	91.9
Fed-MedLoRA2026.01	80.9	89.3
Centralized fine-tuning2026.01	80.7	90.7
Single site fine-tuning2026.01	80.3	91.8
Fed-MedLoRA+2026.01	80.3	90
Fed-MedLoRA+2026.01	80.2	88.8
Fed-MedLoRA2026.01	79.9	86.5
Single site fine-tuning2026.01	79.7	90.2
Bio_ClinicalBERT Single site2026.01	78.2	84.8
DeepSeek-R1-Distill-8B Single site2026.01	77.8	89.5
LLaMA3-8B Single site2026.01	77.5	87.6
Single site fine-tuning2026.01	75.3	86.7
Zero-shot2026.01	55.6	81.5
GPT-4o Zero-shot2026.01	55.6	81.5
Zero-shot2026.01	34.5	50.3
LLaMA3-8B Zero-shot2026.01	34.5	50.3
Zero-shot2026.01	29.1	43.8
DeepSeek-R1-Distill-8B Zero-shot2026.01	29.1	43.8