Medical Natural Language Inference
The MedNLI dataset consists of the sentence pairs developed by Physicians from the Past Medical History section of MIMIC-III clinical notes annotated for Definitely True, Maybe True and Definitely False. The dataset contains 11,232 training, 1,395 development and 1,422 test instances. This provides a natural language inference task (NLI) grounded in the medical history of patients.
Source: MT-Clinical BERT: Scaling Clinical Information Extraction with Multitask Learning
Image Source: https://arxiv.org/abs/1904.02181
Variants: MedNLI
This dataset is used in 2 benchmarks:
| Task | Model | Paper | Date |
|---|---|---|---|
| Natural Language Inference | ClinicalMosaic | Patient Trajectory Prediction: Integrating Clinical … | 2025-02-25 |
| Natural Language Inference | BiomedGPT-B | BiomedGPT: A Generalist Vision-Language Foundation … | 2023-05-26 |
| Few-Shot Learning | CoT-T5-11B (1024 Shot) | The CoT Collection: Improving Zero-shot … | 2023-05-23 |
| Natural Language Inference | SciFive-large | SciFive: a text-to-text transformer model … | 2021-05-28 |
| Natural Language Inference | CharacterBERT (base, medical) | CharacterBERT: Reconciling ELMo and BERT … | 2020-10-20 |
| Natural Language Inference | NCBI_BERT(base) (P+M) | Transfer Learning in Biomedical Natural … | 2019-06-13 |
Recent papers with results on this dataset: