Eric K. Oermann

Sully F. Chen, Robert J. Steele, Glen M. Hocky et al.

The transformer architecture has revolutionized bioinformatics and driven progress in the understanding and prediction of the properties of biomolecules. To date, most biosequence transformers have been trained on single-omic data - either proteins or nucleic acids - and have seen incredible success in downstream tasks in each domain, with particularly noteworthy breakthroughs in protein structural modeling. However, single-omic pretraining limits the ability of these models to capture cross-modal interactions. Here we present OmniBioTE, the largest open-source multi-omic model trained on over 250 billion tokens of mixed protein and nucleic acid data. We show that despite only being trained on unlabeled sequence data, OmniBioTE learns joint representations mapping genes to their corresponding protein sequences. We further demonstrate that OmniBioTE achieves state-of-the-art results predicting the change in Gibbs free energy ({ΔG}) of the binding interaction between a given nucleic acid and protein. Remarkably, we show that multi-omic biosequence transformers emergently learn useful structural information without any a priori structural training, allowing us to predict which protein residues are most involved in the protein-nucleic acid binding interaction. Compared to single-omic controls trained with identical compute, OmniBioTE also demonstrates superior performance-per-FLOP across both multi-omic and single-omic benchmarks. Together, these results highlight the power of a unified modeling approach for biological sequences and establish OmniBioTE as a foundation model for multi-omic discovery.

Grace Yang, Ming Cao, Lavender Y. Jiang et al.

Traditional evaluation metrics for classification in natural language processing such as accuracy and area under the curve fail to differentiate between models with different predictive behaviors despite their similar performance metrics. We introduce sensitivity score, a metric that scrutinizes models' behaviors at the vocabulary level to provide insights into disparities in their decision-making logic. We assess the sensitivity score on a set of representative words in the test set using two classifiers trained for hospital readmission classification with similar performance statistics. Our experiments compare the decision-making logic of clinicians and classifiers based on rank correlations of sensitivity scores. The results indicate that the language model's sensitivity score aligns better with the professionals than the xgboost classifier on tf-idf embeddings, which suggests that xgboost uses some spurious features. Overall, this metric offers a novel perspective on assessing models' robustness by quantifying their discrepancy with professional opinions. Our code is available on GitHub (https://github.com/nyuolab/Model_Sensitivity).

Julián N. Acosta, Xiaoman Zhang, Siddhant Dogra et al.

We present Head CT Ontology Normalized Evaluation (HeadCT-ONE), a metric for evaluating head CT report generation through ontology-normalized entity and relation extraction. HeadCT-ONE enhances current information extraction derived metrics (such as RadGraph F1) by implementing entity normalization through domain-specific ontologies, addressing radiological language variability. HeadCT-ONE compares normalized entities and relations, allowing for controllable weighting of different entity types or specific entities. Through experiments on head CT reports from three health systems, we show that HeadCT-ONE's normalization and weighting approach improves the capture of semantically equivalent reports, better distinguishes between normal and abnormal reports, and aligns with radiologists' assessment of clinically significant errors, while offering flexibility to prioritize specific aspects of report content. Our results demonstrate how HeadCT-ONE enables more flexible, controllable, and granular automated evaluation of head CT reports.

Krithik Vishwanath, Mrigayu Ghosh, Anton Alyakin et al.

Specialized clinical AI assistants are rapidly entering medical practice, often framed as safer or more reliable than general-purpose large language models (LLMs). Yet, unlike frontier models, these clinical tools are rarely subjected to independent, quantitative evaluation, creating a critical evidence gap despite their growing influence on diagnosis, triage, and guideline interpretation. We assessed two widely deployed clinical AI systems (OpenEvidence and UpToDate Expert AI) against three state-of-the-art generalist LLMs (GPT-5, Gemini 3 Pro, and Claude Sonnet 4.5) using a 1,000-item mini-benchmark combining MedQA (medical knowledge) and HealthBench (clinician-alignment) tasks. Generalist models consistently outperformed clinical tools, with GPT-5 achieving the highest scores, while OpenEvidence and UpToDate demonstrated deficits in completeness, communication quality, context awareness, and systems-based safety reasoning. These findings reveal that tools marketed for clinical decision support may often lag behind frontier LLMs, underscoring the urgent need for transparent, independent evaluation before deployment in patient-facing workflows.

Hongyi Zheng, Yixin Zhu, Lavender Yao Jiang et al.

Recent advances in large language models have led to renewed interest in natural language processing in healthcare using the free text of clinical notes. One distinguishing characteristic of clinical notes is their long time span over multiple long documents. The unique structure of clinical notes creates a new design choice: when the context length for a language model predictor is limited, which part of clinical notes should we choose as the input? Existing studies either choose the inputs with domain knowledge or simply truncate them. We propose a framework to analyze the sections with high predictive power. Using MIMIC-III, we show that: 1) predictive power distribution is different between nursing notes and discharge notes and 2) combining different types of notes could improve performance when the context length is large. Our findings suggest that a carefully selected sampling function could enable more efficient information extraction from clinical notes.

Krithik Vishwanath, Anton Alyakin, Mrigayu Ghosh et al.

The Congress of Neurological Surgeons Self-Assessment for Neurological Surgeons (CNS-SANS) questions are widely used by neurosurgical residents to prepare for written board examinations. Recently, these questions have also served as benchmarks for evaluating large language models' (LLMs) neurosurgical knowledge. This study aims to assess the performance of state-of-the-art LLMs on neurosurgery board-like questions and to evaluate their robustness to the inclusion of distractor statements. A comprehensive evaluation was conducted using 28 large language models. These models were tested on 2,904 neurosurgery board examination questions derived from the CNS-SANS. Additionally, the study introduced a distraction framework to assess the fragility of these models. The framework incorporated simple, irrelevant distractor statements containing polysemous words with clinical meanings used in non-clinical contexts to determine the extent to which such distractions degrade model performance on standard medical benchmarks. 6 of the 28 tested LLMs achieved board-passing outcomes, with the top-performing models scoring over 15.7% above the passing threshold. When exposed to distractions, accuracy across various model architectures was significantly reduced-by as much as 20.4%-with one model failing that had previously passed. Both general-purpose and medical open-source models experienced greater performance declines compared to proprietary variants when subjected to the added distractors. While current LLMs demonstrate an impressive ability to answer neurosurgery board-like exam questions, their performance is markedly vulnerable to extraneous, distracting information. These findings underscore the critical need for developing novel mitigation strategies aimed at bolstering LLM resilience against in-text distractions, particularly for safe and effective clinical deployment.

Yanbing Chen, Ruilin Wang, Zihao Yang et al.

Packing and shuffling tokens is a common practice in training auto-regressive language models (LMs) to prevent overfitting and improve efficiency. Typically documents are concatenated to chunks of maximum sequence length (MSL) and then shuffled. However setting the atom size, the length for each data chunk accompanied by random shuffling, to MSL may lead to contextual incoherence due to tokens from different documents being packed into the same chunk. An alternative approach is to utilize padding, another common data packing strategy, to avoid contextual incoherence by only including one document in each shuffled chunk. To optimize both packing strategies (concatenation vs padding), we investigated the optimal atom size for shuffling and compared their performance and efficiency. We found that matching atom size to MSL optimizes performance for both packing methods (concatenation and padding), and padding yields lower final perplexity (higher performance) than concatenation at the cost of more training steps and lower compute efficiency. This trade-off informs the choice of packing methods in training language models.

Salman Rahman, Lavender Yao Jiang, Saadia Gabriel et al.

Advances in large language models (LLMs) provide new opportunities in healthcare for improved patient care, clinical decision-making, and enhancement of physician and administrator workflows. However, the potential of these models importantly depends on their ability to generalize effectively across clinical environments and populations, a challenge often underestimated in early development. To better understand reasons for these challenges and inform mitigation approaches, we evaluated ClinicLLM, an LLM trained on [HOSPITAL]'s clinical notes, analyzing its performance on 30-day all-cause readmission prediction focusing on variability across hospitals and patient characteristics. We found poorer generalization particularly in hospitals with fewer samples, among patients with government and unspecified insurance, the elderly, and those with high comorbidities. To understand reasons for lack of generalization, we investigated sample sizes for fine-tuning, note content (number of words per note), patient characteristics (comorbidity level, age, insurance type, borough), and health system aspects (hospital, all-cause 30-day readmission, and mortality rates). We used descriptive statistics and supervised classification to identify features. We found that, along with sample size, patient age, number of comorbidities, and the number of words in notes are all important factors related to generalization. Finally, we compared local fine-tuning (hospital specific), instance-based augmented fine-tuning and cluster-based fine-tuning for improving generalization. Among these, local fine-tuning proved most effective, increasing AUC by 0.25% to 11.74% (most helpful in settings with limited data). Overall, this study provides new insights for enhancing the deployment of large language models in the societally important domain of healthcare, and improving their performance for broader populations.

Gabriel R. Rosenbaum, Lavender Yao Jiang, Ivaxi Sheth et al.

Large language models (LLMs) have recently emerged as powerful tools, finding many medical applications. LLMs' ability to coalesce vast amounts of information from many sources to generate a response-a process similar to that of a human expert-has led many to see potential in deploying LLMs for clinical use. However, medicine is a setting where accurate reasoning is paramount. Many researchers are questioning the effectiveness of multiple choice question answering (MCQA) benchmarks, frequently used to test LLMs. Researchers and clinicians alike must have complete confidence in LLMs' abilities for them to be deployed in a medical setting. To address this need for understanding, we introduce a knowledge graph (KG)-based method to evaluate the biomedical reasoning abilities of LLMs. Essentially, we map how LLMs link medical concepts in order to better understand how they reason. We test GPT-4, Llama3-70b, and PalmyraMed-70b, a specialized medical model. We enlist a panel of medical students to review a total of 60 LLM-generated graphs and compare these graphs to BIOS, a large biomedical KG. We observe GPT-4 to perform best in our human review but worst in our ground truth comparison; vice-versa with PalmyraMed, the medical model. Our work provides a means of visualizing the medical reasoning pathways of LLMs so they can be implemented in clinical settings safely and effectively.

Lavender Y. Jiang, Angelica Chen, Xu Han et al.

Hospitals and healthcare systems rely on operational decisions that determine patient flow, cost, and quality of care. Despite strong performance on medical knowledge and conversational benchmarks, foundation models trained on general text may lack the specialized knowledge required for these operational decisions. We introduce Lang1, a family of models (100M-7B parameters) pretrained on a specialized corpus blending 80B clinical tokens from NYU Langone Health's EHRs and 627B tokens from the internet. To rigorously evaluate Lang1 in real-world settings, we developed the REalistic Medical Evaluation (ReMedE), a benchmark derived from 668,331 EHR notes that evaluates five critical tasks: 30-day readmission prediction, 30-day mortality prediction, length of stay, comorbidity coding, and predicting insurance claims denial. In zero-shot settings, both general-purpose and specialized models underperform on four of five tasks (36.6%-71.7% AUROC), with mortality prediction being an exception. After finetuning, Lang1-1B outperforms finetuned generalist models up to 70x larger and zero-shot models up to 671x larger, improving AUROC by 3.64%-6.75% and 1.66%-23.66% respectively. We also observed cross-task scaling with joint finetuning on multiple tasks leading to improvement on other tasks. Lang1-1B effectively transfers to out-of-distribution settings, including other clinical tasks and an external health system. Our findings suggest that predictive capabilities for hospital operations require explicit supervised finetuning, and that this finetuning process is made more efficient by in-domain pretraining on EHR. Our findings support the emerging view that specialized LLMs can compete with generalist models in specialized tasks, and show that effective healthcare systems AI requires the combination of in-domain pretraining, supervised finetuning, and real-world evaluation beyond proxy benchmarks.

Ioana Marinescu, Kyunghyun Cho, Eric Karl Oermann

In-context learning (ICL) is the ability of a large language model (LLM) to learn a new task from a few demonstrations presented as part of the context. Past studies have attributed a large portion of the success of ICL to the way these in-context demonstrations are represented, particularly to how labels are represented in classification tasks. On the other hand, observations of the learning capacity of ICL (i.e., the extent to which more in-context demonstrations can lead to higher performance) have been mixed, and ICL is often thought to occur only under specific conditions. The interaction between these two aspects in ICL, representation and learning, has not been studied in depth until now. We hypothesize that they are largely independent of one another, such that the representation of demonstrations determines the baseline accuracy of ICL, while learning from additional demonstrations improves only on top of this baseline. We validate this hypothesis by developing an optimization algorithm that can enumerate a spectrum of possible label sets (representations) varying in semantic relevance. We then perform ICL with varying numbers of in-context demonstrations for each of these label sets. We observed that learning happens regardless of the quality of the label set itself, although its efficiency, measured by the slope of improvement over in-context demonstrations, is conditioned on both the label set quality and the parameter count of the underlying language model. Despite the emergence of learning, the relative quality (accuracy) of the choice of a label set (representation) is largely maintained throughout learning, confirming our hypothesis and implying their orthogonality. Our work reveals a previously underexplored aspect of ICL: the independent effects of learning from demonstrations and their representations on ICL performance.

Chi Hang, Ruiqi Deng, Lavender Yao Jiang et al.

Clinical measurements such as blood pressures and respiration rates are critical in diagnosing and monitoring patient outcomes. It is an important component of biomedical data, which can be used to train transformer-based language models (LMs) for improving healthcare delivery. It is, however, unclear whether LMs can effectively interpret and use clinical measurements. We investigate two questions: First, can LMs effectively leverage clinical measurements to answer related medical questions? Second, how to enhance an LM's performance on medical question-answering (QA) tasks that involve measurements? We performed a case study on blood pressure readings (BPs), a vital sign routinely monitored by medical professionals. We evaluated the performance of four LMs: BERT, BioBERT, MedAlpaca, and GPT-3.5, on our newly developed dataset, BPQA (Blood Pressure Question Answering). BPQA contains $100$ medical QA pairs that were verified by medical students and designed to rely on BPs . We found that GPT-3.5 and MedAlpaca (larger and medium sized LMs) benefit more from the inclusion of BPs than BERT and BioBERT (small sized LMs). Further, augmenting measurements with labels improves the performance of BioBERT and Medalpaca (domain specific LMs), suggesting that retrieval may be useful for improving domain-specific LMs.

Krithik Vishwanath, Jaden Stryker, Anton Alyakin et al.

Language models (LMs) have demonstrated expert-level reasoning and recall abilities in medicine. However, computational costs and privacy concerns are mounting barriers to wide-scale implementation. To address these significant limitations, we introduce a parsimonious adaptation of phi-3-mini, MedMobile, a 3.8 billion parameter LM capable of running on a mobile device, for medical applications. We perform a careful set of pipeline additions and demonstrate that chain of thought, ensembling, and fine-tuning lead to the greatest performance gains, while unexpectedly retrieval augmented generation fails to demonstrate significant improvements. We evaluate the efficiency of our pipeline on the MultiMedQA and MedBullets. We demonstrate that MedMobile scores 75.7% on the MedQA (USMLE), surpassing the passing mark for licensed physicians (~60%) and rivaling scores of models 100 times its size. Across the entirety of the MultiMedQA, MedMobile achieves SOTA performance for models with less than 5B parameters and represents the smallest model to pass the MedQA (USMLE). MedMobile holds promise to democratize access to language models in medicine, bolstering lower compute needs and fast inference speeds. With the ability to combat the biggest barriers to entry for language models in medicine, we hope that MedMobile is a critical step forward in developing clinically relevant language models.

Yifan Li, Garrett Yoon, Mustafa Nasir-Moin et al.

Numerous COVID-19 clinical decision support systems have been developed. However many of these systems do not have the merit for validity due to methodological shortcomings including algorithmic bias. Methods Logistic regression models were created to predict COVID-19 mortality, ventilator status and inpatient status using a real-world dataset consisting of four hospitals in New York City and analyzed for biases against race, gender and age. Simple thresholding adjustments were applied in the training process to establish more equitable models. Results Compared to the naively trained models, the calibrated models showed a 57% decrease in the number of biased trials, while predictive performance, measured by area under the receiver/operating curve (AUC), remained unchanged. After calibration, the average sensitivity of the predictive models increased from 0.527 to 0.955. Conclusion We demonstrate that naively training and deploying machine learning models on real world data for predictive analytics of COVID-19 has a high risk of bias. Simple implemented adjustments or calibrations during model training can lead to substantial and sustained gains in fairness on subsequent deployment.

Brian Murphy, Mustafa Nasir-Moin, Grace von Oiste et al.

The prognosis for patients with epithelial ovarian cancer remains dismal despite improvements in survival for other cancers. Treatment involves multiple lines of chemotherapy and becomes increasingly heterogeneous after first-line therapy. Reinforcement learning with real-world outcomes data has the potential to identify novel treatment strategies to improve overall survival. We design a reinforcement learning environment to model epithelial ovarian cancer treatment trajectories and use model free reinforcement learning to investigate therapeutic regimens for simulated patients.

Annika Brundyn, Jesse Swanson, Kyunghyun Cho et al.

We introduce the task of stereo video reconstruction or, equivalently, 2D-to-3D video conversion for minimally invasive surgical video. We design and implement a series of end-to-end U-Net-based solutions for this task by varying the input (single frame vs. multiple consecutive frames), loss function (MSE, MAE, or perceptual losses), and network architecture. We evaluate these solutions by surveying ten experts - surgeons who routinely perform endoscopic surgery. We run two separate reader studies: one evaluating individual frames and the other evaluating fully reconstructed 3D video played on a VR headset. In the first reader study, a variant of the U-Net that takes as input multiple consecutive video frames and outputs the missing view performs best. We draw two conclusions from this outcome. First, motion information coming from multiple past frames is crucial in recreating stereo vision. Second, the proposed U-Net variant can indeed exploit such motion information for solving this task. The result from the second study further confirms the effectiveness of the proposed U-Net variant. The surgeons reported that they could successfully perceive depth from the reconstructed 3D video clips. They also expressed a clear preference for the reconstructed 3D video over the original 2D video. These two reader studies strongly support the usefulness of the proposed task of stereo reconstruction for minimally invasive surgical video and indicate that deep learning is a promising approach to this task. Finally, we identify two automatic metrics, LPIPS and DISTS, that are strongly correlated with expert judgement and that could serve as proxies for the latter in future studies.

Aahlad Puli, Lily H. Zhang, Eric K. Oermann et al.

In many prediction problems, spurious correlations are induced by a changing relationship between the label and a nuisance variable that is also correlated with the covariates. For example, in classifying animals in natural images, the background, which is a nuisance, can predict the type of animal. This nuisance-label relationship does not always hold, and the performance of a model trained under one such relationship may be poor on data with a different nuisance-label relationship. To build predictive models that perform well regardless of the nuisance-label relationship, we develop Nuisance-Randomized Distillation (NURD). We introduce the nuisance-randomized distribution, a distribution where the nuisance and the label are independent. Under this distribution, we define the set of representations such that conditioning on any member, the nuisance and the label remain independent. We prove that the representations in this set always perform better than chance, while representations outside of this set may not. NURD finds a representation from this set that is most informative of the label under the nuisance-randomized distribution, and we prove that this representation achieves the highest performance regardless of the nuisance-label relationship. We evaluate NURD on several tasks including chest X-ray classification where, using non-lung patches as the nuisance, NURD produces models that predict pneumonia under strong spurious correlations.

Daniel Ranti, Katie Hanss, Shan Zhao et al.

The purpose of this study is to analyze the efficacy of transfer learning techniques and transformer-based models as applied to medical natural language processing (NLP) tasks, specifically radiological text classification. We used 1,977 labeled head CT reports, from a corpus of 96,303 total reports, to evaluate the efficacy of pretraining using general domain corpora and a combined general and medical domain corpus with a bidirectional representations from transformers (BERT) model for the purpose of radiological text classification. Model performance was benchmarked to a logistic regression using bag-of-words vectorization and a long short-term memory (LSTM) multi-label multi-class classification model, and compared to the published literature in medical text classification. The BERT models using either set of pretrained checkpoints outperformed the logistic regression model, achieving sample-weighted average F1-scores of 0.87 and 0.87 for the general domain model and the combined general and biomedical-domain model. General text transfer learning may be a viable technique to generate state-of-the-art results within medical NLP tasks on radiological corpora, outperforming other deep models such as LSTMs. The efficacy of pretraining and transformer-based models could serve to facilitate the creation of groundbreaking NLP models in the uniquely challenging data environment of medical text.

John R. Zech, Marcus A. Badgeley, Manway Liu et al.

Early results in using convolutional neural networks (CNNs) on x-rays to diagnose disease have been promising, but it has not yet been shown that models trained on x-rays from one hospital or one group of hospitals will work equally well at different hospitals. Before these tools are used for computer-aided diagnosis in real-world clinical settings, we must verify their ability to generalize across a variety of hospital systems. A cross-sectional design was used to train and evaluate pneumonia screening CNNs on 158,323 chest x-rays from NIH (n=112,120 from 30,805 patients), Mount Sinai (42,396 from 12,904 patients), and Indiana (n=3,807 from 3,683 patients). In 3 / 5 natural comparisons, performance on chest x-rays from outside hospitals was significantly lower than on held-out x-rays from the original hospital systems. CNNs were able to detect where an x-ray was acquired (hospital system, hospital department) with extremely high accuracy and calibrate predictions accordingly. The performance of CNNs in diagnosing diseases on x-rays may reflect not only their ability to identify disease-specific imaging findings on x-rays, but also their ability to exploit confounding information. Estimates of CNN performance based on test data from hospital systems used for model training may overstate their likely real-world performance.

Varun Arvind, Anthony Costa, Marcus Badgeley et al.

3D shape models that directly classify objects from 3D information have become more widely implementable. Current state of the art models rely on deep convolutional and inception models that are resource intensive. Residual neural networks have been demonstrated to be easier to optimize and do not suffer from vanishing/exploding gradients observed in deep networks. Here we implement a residual neural network for 3D object classification of the 3D Princeton ModelNet dataset. Further, we show that widening network layers dramatically improves accuracy in shallow residual nets, and residual neural networks perform comparable to state-of-the-art 3D shape net models, and we show that widening network layers improves classification accuracy. We provide extensive training and architecture parameters providing a better understanding of available network architectures for use in 3D object classification.