Named Entity Recognition on BC5CDR
93.5F1 ScoreSOTA
Evaluation Results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| SOTASupervision Source=Hand-labeled2020.08 | 93.5 | — | — | — | — | |
| Fully supervised BioBERT (FS)Supervision Source=Hand-labeled2020.08 | 92.4 | — | — | — | — | |
| InstructUIE2023.04 | 91.9 | — | — | — | — | |
| Weakly supervised BioBERT (WS)Supervision Source=Ontologies + Task-specific Rules, LFs=312020.08 | 91.1 | — | — | — | — | |
| CL-L2Evaluation Setting=W/ CONTEXT2021.05 | 90.99 | — | — | — | — | |
| CL-KLEvaluation Setting=W/ CONTEXT2021.05 | 90.93 | — | — | — | — | |
| SpanDec (RoBERTa-L)Params=0.35B, Throughput=118.7 (1×)2026.04 | 90.8 | — | — | — | — | |
| W/ CONTEXT (Baseline)Evaluation Setting=W/ CONTEXT2021.05 | 90.76 | — | — | — | — | |
| CL-KLEvaluation Setting=W/O CONTEXT2021.05 | 90.73 | — | — | — | — | |
| CL-L2Evaluation Setting=W/O CONTEXT2021.05 | 90.7 | — | — | — | — | |
| W/O CONTEXT (Baseline)Evaluation Setting=W/O CONTEXT2021.05 | 90.52 | — | — | — | — | |
| GNEREvaluation Protocol=Supervised2026.04 | 90.3 | — | — | — | — | |
| Nooralahzadeh et al.2021.05 | 89.93 | — | — | — | — | |
| Spark-BiomedicalImplementation=Spark-Biomedical2020.11 | 89.73 | — | — | — | — | |
| InstructUIEBackbone=11B FlanT52023.04 | 89.59 | — | — | — | — | |
| Bio-Flair2021.05 | 89.42 | — | — | — | — | |
| UniversalNERParameters=7B2026.04 | 89.34 | — | — | — | — | |
| KnowCoderParameters=7B2026.04 | 89.3 | — | — | — | — | |
| UniNER-7BEvaluation Protocol=Supervised2026.04 | 89.3 | — | — | — | — | |
| KnowCoder-7BEvaluation Protocol=Supervised2026.04 | 89.3 | — | — | — | — | |
| Label Model (LM)Supervision Source=Ontologies + Task-specific Rules, LFs=312020.08 | 89.2 | — | — | — | — | |
| LUKEEvaluation Setting=W/O CONTEXT2021.05 | 89.18 | — | — | — | — | |
| InstructUIEParameters=11B2026.04 | 89.02 | — | — | — | — | |
| InstructUIEEvaluation Protocol=Supervised2026.04 | 89 | — | — | — | — | |
| InstructUIEParams=13B, Throughput=< 5.8 (0.05×)2026.04 | 89 | — | — | — | — | |
| UniNERParams=7B, Throughput=< 6.4 (0.05×)2026.04 | 89 | — | — | — | — | |
| ProUIEParameters=4B2026.04 | 88.93 | — | — | — | — | |
| Supervised BERT-NERSupervision Type=Fully Supervised, Pre-trained Model=SciBERT2021.05 | 88.81 | 87.12 | 90.57 | — | — | |
| GLiNEREvaluation Protocol=Supervised2026.04 | 88.7 | — | — | — | — | |
| GLiNERParams=0.35B, Throughput=76.22026.04 | 88.7 | — | — | — | — | |
| B2NEREvaluation Protocol=Supervised2024.11 | 88.52 | — | — | — | — | |
| Weakly supervised BioBERT (WS)Supervision Source=Ontologies, LFs=222020.08 | 88.5 | — | — | — | — | |
| B2NEREvaluation Protocol=Supervised2026.04 | 88.5 | — | — | — | — | |
| KnowCoder-XEvaluation Protocol=Supervised2024.11 | 88.46 | — | — | — | — | |
| GollIEParameters=34B2026.04 | 88.4 | — | — | — | — | |
| Spark-GloVe6BImplementation=Spark-GloVe6B2020.11 | 88.32 | — | — | — | — | |
| Stanza2020.11 | 88.08 | — | — | — | — | |
| IEPILEEvaluation Protocol=Supervised2024.11 | 88.07 | — | — | — | — | |
| Label Model (LM)Supervision Source=Ontologies, LFs=222020.08 | 88 | — | — | — | — | |
| GollIEParameters=13B2026.04 | 87.9 | — | — | — | — | |
| Bio-BERT2021.05 | 87.7 | — | — | — | — | |
| best consensusSupervision Type=Oracle2021.05 | 87.58 | 100 | 77.89 | — | — | |
| GollIEParameters=7B2026.04 | 87.5 | — | — | — | — | |
| RDANERtraining_data=full2021.01 | 87.38 | — | — | — | — | |
| SOTASupervision Source=Hand-labeled2020.08 | 87.2 | — | — | — | — | |
| SpERTtraining_data=full, paradigm=multi-task learning2021.01 | 86.65 | — | — | — | — | |
| GPT (Moderation)Approach=Moderation (M), Iterations=12026.05 | 86 | 92 | 81 | — | 1,737 | |
| DyGIE++training_data=full, paradigm=multi-task learning2021.01 | 85.44 | — | — | — | — | |
| BERT-baseEvaluation Protocol=Supervised2026.04 | 85.3 | — | — | — | — | |
| BERTEvaluation Protocol=Supervised2024.11 | 85.28 | — | — | — | — | |
| Bert-baseBackbone=Bert-base2023.04 | 85.28 | — | — | — | — | |
| BERT-base2026.04 | 85.28 | — | — | — | — | |
| CHMM-ALTSupervision Type=Distant Supervision, Pre-trained Model=SciBERT2021.05 | 85.12 | 84.97 | 85.28 | — | — | |
| GPT (Guideline)Approach=Guideline (G), Iterations=12026.05 | 85 | 89 | 81 | — | 1,735 | |
| AutoNERtraining_data=full, paradigm=distantly supervised learning2021.01 | 84.8 | — | — | — | — | |
| Fully supervised BioBERT (FS)Supervision Source=Hand-labeled2020.08 | 84.5 | — | — | — | — | |
| CHMM + BERT-NERSupervision Type=Distant Supervision, Pre-trained Model=SciBERT2021.05 | 84.33 | 85.58 | 83.12 | — | — | |
| SwellSharktraining_data=full, paradigm=distantly supervised learning2021.01 | 84.23 | — | — | — | — | |
| SwellShark (noun-phrase)Supervision Type=Distant Supervision, Label Denoiser Status=Active2021.05 | 84.23 | 84.98 | 83.49 | — | — | |
| SwellShark (hand-tuned)Supervision Type=Distant Supervision, Label Denoiser Status=Active2021.05 | 84.21 | 86.11 | 82.39 | — | — | |
| YAYI-UIEEvaluation Protocol=Supervised2024.11 | 83.67 | — | — | — | — | |
| BOND-MVSupervision Type=Distant Supervision, Label Denoiser Status=Active2021.05 | 83.18 | 82.9 | 83.49 | — | — | |
| Linked HMMSupervision Type=Distant Supervision, Label Denoiser Status=Active2021.05 | 82.96 | 82.65 | 83.28 | — | — | |
| CHMMSupervision Type=Unsupervised label denoiser2021.05 | 82.39 | 89.93 | 76.02 | — | — | |
| SnorkelSupervision Type=Distant Supervision, Label Denoiser Status=Active2021.05 | 82.24 | 80.23 | 84.35 | — | — | |
| AutoNERSupervision Type=Distant Supervision, Label Denoiser Status=Active2021.05 | 82.13 | 83.23 | 81.06 | — | — | |
| Majority Vote (MV)Supervision Source=Ontologies + Task-specific Rules, LFs=312020.08 | 81.1 | — | — | — | — | |
| Majority VotingSupervision Type=Unsupervised label denoiser2021.05 | 80.73 | 83.79 | 77.88 | — | — | |
| HMMSupervision Type=Unsupervised label denoiser2021.05 | 80.57 | 88.75 | 73.76 | — | — | |
| GPT (Simple prompting)Approach=Simple prompting (S), Iterations=12026.05 | 80 | 84 | 78 | — | 1,664 | |
| Weakly supervised BioBERT (WS)Supervision Source=Ontologies + Task-specific Rules, LFs=222020.08 | 79.9 | — | — | — | — | |
| Majority Vote (MV)Supervision Source=Ontologies, LFs=222020.08 | 79.8 | — | — | — | — | |
| Label Model (LM)Supervision Source=Ontologies + Task-specific Rules, LFs=222020.08 | 79.8 | — | — | — | — | |
| CHMM-i.i.d.Supervision Type=Unsupervised label denoiser2021.05 | 79.37 | 85.68 | 73.92 | — | — | |
| Label Model (LM)Supervision Source=Ontologies, LFs=162020.08 | 78.9 | — | — | — | — | |
| DiZiNERZero-shot=true2026.04 | 78.9 | — | — | — | — | |
| DiZiNEREvaluation Protocol=Zero-shot2026.04 | 78.9 | — | — | — | — | |
| Weakly supervised BioBERT (WS)Supervision Source=Ontologies, LFs=162020.08 | 78.3 | — | — | — | — | |
| MACEEnsemble=true2026.04 | 77.6 | — | — | — | — | |
| GLADEnsemble=true2026.04 | 77.5 | — | — | — | — | |
| MVEnsemble=true2026.04 | 77.1 | — | — | — | — | |
| Gemini (Moderation)Approach=Moderation (M), Iterations=12026.05 | 77 | 86 | 70 | — | 1,503 | |
| Majority Vote (MV)Supervision Source=Ontologies + Task-specific Rules, LFs=222020.08 | 76.4 | — | — | — | — | |
| Gemini (Guideline)Approach=Guideline (G), Iterations=12026.05 | 76 | 84 | 68 | — | 1,469 | |
| Majority Vote (MV)Supervision Source=Ontologies, LFs=162020.08 | 74.7 | — | — | — | — | |
| JPT-4Bzero-shot=true2026.04 | 70.4 | — | — | — | — | |
| gliner_large-v2.5Evaluation Protocol=Zero-shot2026.02 | 68.7 | — | — | — | — | |
| UniNER-7Bzero-shot=true2026.04 | 68 | — | — | — | — | |
| Prior Best ZSZero-shot=true2026.04 | 68 | — | — | — | — | |
| UniNER-7BEvaluation Protocol=Zero-shot2026.04 | 68 | — | — | — | — | |
| Gemini (Simple prompting)Approach=Simple prompting (S), Iterations=12026.05 | 68 | 74 | 63 | — | 1,359 | |
| GLiNER-Lzero-shot=true2026.04 | 66.4 | — | — | — | — | |
| GLiNEREvaluation Protocol=Zero-shot2026.04 | 66.4 | — | — | — | — | |
| gliner_small-v2.5Evaluation Protocol=Zero-shot2026.02 | 65.9 | — | — | — | — | |
| gliner_medium-v2.5Evaluation Protocol=Zero-shot2026.02 | 65.9 | — | — | — | — | |
| DSEnsemble=true2026.04 | 65.9 | — | — | — | — | |
| DeepSeek (Moderation)Approach=Moderation (M), Iterations=12026.05 | 65 | 86 | 52 | — | 1,119 | |
| DeepSeek (Guideline)Approach=Guideline (G), Iterations=12026.05 | 64 | 89 | 50 | — | 1,072 | |
| GPT-5 miniZero-shot=true2026.04 | 62.8 | — | — | — | — | |
| GPT-5 miniEvaluation Protocol=Zero-shot, Note=supervisor model2026.04 | 62.8 | — | — | — | — |