This study presents a 2D map of biomedical research based on abstracts from 21 million English articles in the PubMed database. The map, created using the large language model PubMedBERT and t-SNE, highlights various publishing issues such as gender bias and fraudulent research. Key findings include: 1. **COVID-19 Literature**: The COVID-19 literature is uniquely isolated, with most papers grouped together in one cluster, indicating a significant impact on the scientific literature. 2. **Neuroscience Evolution**: The map reveals shifting topics and trends within neuroscience, showing a bimodal separation between cellular and behavioral neuroscience. 3. **Machine Learning Adoption**: The uptake of machine learning varies across medical disciplines, with radiology leading the way, followed by psychiatry and neurology. 4. **Gender Imbalance**: The map illustrates the gender imbalance in academic authorship, with female authors being more prevalent in certain fields like nursing and education. 5. **Retraction hotspots**: Retracted papers are concentrated in specific areas, particularly those related to cancer drugs, marker genes, and microRNA, suggesting the presence of paper mills. The interactive web version of the map allows users to explore the literature, navigate the atlas, and search by title, journal, or author names. The study demonstrates the utility of 2D visualizations in uncovering detailed insights into the biomedical research landscape.This study presents a 2D map of biomedical research based on abstracts from 21 million English articles in the PubMed database. The map, created using the large language model PubMedBERT and t-SNE, highlights various publishing issues such as gender bias and fraudulent research. Key findings include: 1. **COVID-19 Literature**: The COVID-19 literature is uniquely isolated, with most papers grouped together in one cluster, indicating a significant impact on the scientific literature. 2. **Neuroscience Evolution**: The map reveals shifting topics and trends within neuroscience, showing a bimodal separation between cellular and behavioral neuroscience. 3. **Machine Learning Adoption**: The uptake of machine learning varies across medical disciplines, with radiology leading the way, followed by psychiatry and neurology. 4. **Gender Imbalance**: The map illustrates the gender imbalance in academic authorship, with female authors being more prevalent in certain fields like nursing and education. 5. **Retraction hotspots**: Retracted papers are concentrated in specific areas, particularly those related to cancer drugs, marker genes, and microRNA, suggesting the presence of paper mills. The interactive web version of the map allows users to explore the literature, navigate the atlas, and search by title, journal, or author names. The study demonstrates the utility of 2D visualizations in uncovering detailed insights into the biomedical research landscape.