Shadab Khan

Clément Christophe, Praveen K Kanithi, Tathagata Raha et al.

Med42-v2 introduces a suite of clinical large language models (LLMs) designed to address the limitations of generic models in healthcare settings. These models are built on Llama3 architecture and fine-tuned using specialized clinical data. They underwent multi-stage preference alignment to effectively respond to natural prompts. While generic models are often preference-aligned to avoid answering clinical queries as a precaution, Med42-v2 is specifically trained to overcome this limitation, enabling its use in clinical settings. Med42-v2 models demonstrate superior performance compared to the original Llama3 models in both 8B and 70B parameter configurations and GPT-4 across various medical benchmarks. These LLMs are developed to understand clinical queries, perform reasoning tasks, and provide valuable assistance in clinical environments. The models are now publicly available at \href{https://huggingface.co/m42-health}{https://huggingface.co/m42-health}.

Praveenkumar Kanithi, Clément Christophe, Marco AF Pimentel et al.

While Large Language Models (LLMs) achieve superhuman performance on standardized medical licensing exams, these static benchmarks have become saturated and increasingly disconnected from the functional requirements of clinical workflows. To bridge the gap between theoretical capability and verified utility, we introduce MEDIC, a comprehensive evaluation framework establishing leading indicators across various clinical dimensions. Beyond standard question-answering, we assess operational capabilities using deterministic execution protocols and a novel Cross-Examination Framework (CEF), which quantifies information fidelity and hallucination rates without reliance on reference texts. Our evaluation across a heterogeneous task suite exposes critical performance trade-offs: we identify a significant knowledge-execution gap, where proficiency in static retrieval does not predict success in operational tasks such as clinical calculation or SQL generation. Furthermore, we observe a divergence between passive safety (refusal) and active safety (error detection), revealing that models fine-tuned for high refusal rates often fail to reliably audit clinical documentation for factual accuracy. These findings demonstrate that no single architecture dominates across all dimensions, highlighting the necessity of a portfolio approach to clinical model deployment. As part of this investigation, we released a public leaderboard on Hugging Face.\footnote{https://huggingface.co/spaces/m42-health/MEDIC-Benchmark}

Marco AF Pimentel, Clément Christophe, Tathagata Raha et al.

As large language models (LLMs) continue to evolve, the need for robust and standardized evaluation benchmarks becomes paramount. Evaluating the performance of these models is a complex challenge that requires careful consideration of various linguistic tasks, model architectures, and benchmarking methodologies. In recent years, various frameworks have emerged as noteworthy contributions to the field, offering comprehensive evaluation tests and benchmarks for assessing the capabilities of LLMs across diverse domains. This paper provides an exploration and critical analysis of some of these evaluation methodologies, shedding light on their strengths, limitations, and impact on advancing the state-of-the-art in natural language processing.

Clément Christophe, Tathagata Raha, Svetlana Maslenkova et al.

Large Language Models (LLMs) have demonstrated significant potential in transforming clinical applications. In this study, we investigate the efficacy of four techniques in adapting LLMs for clinical use-cases: continuous pretraining, instruct fine-tuning, NEFTune, and prompt engineering. We employ these methods on Mistral 7B and Mixtral 8x7B models, leveraging a large-scale clinical pretraining dataset of 50 billion tokens and an instruct fine-tuning dataset of 500 million tokens. Our evaluation across various clinical tasks reveals the impact of each technique. While continuous pretraining beyond 250 billion tokens yields marginal improvements on its own, it establishes a strong foundation for instruct fine-tuning. Notably, NEFTune, designed primarily to enhance generation quality, surprisingly demonstrates additional gains on our benchmark. Complex prompt engineering methods further enhance performance. These findings show the importance of tailoring fine-tuning strategies and exploring innovative techniques to optimize LLM performance in the clinical domain.

Long Phan, Alice Gatti, Ziwen Han et al. · amazon-science, apple-ml

Benchmarks are important tools for tracking the rapid advancements in large language model (LLM) capabilities. However, benchmarks are not keeping pace in difficulty: LLMs now achieve over 90\% accuracy on popular benchmarks like MMLU, limiting informed measurement of state-of-the-art LLM capabilities. In response, we introduce Humanity's Last Exam (HLE), a multi-modal benchmark at the frontier of human knowledge, designed to be the final closed-ended academic benchmark of its kind with broad subject coverage. HLE consists of 2,500 questions across dozens of subjects, including mathematics, humanities, and the natural sciences. HLE is developed globally by subject-matter experts and consists of multiple-choice and short-answer questions suitable for automated grading. Each question has a known solution that is unambiguous and easily verifiable, but cannot be quickly answered via internet retrieval. State-of-the-art LLMs demonstrate low accuracy and calibration on HLE, highlighting a significant gap between current LLM capabilities and the expert human frontier on closed-ended academic questions. To inform research and policymaking upon a clear understanding of model capabilities, we publicly release HLE at https://lastexam.ai.

Prateek Munjal, Nasir Hayat, Munawar Hayat et al.

Active learning (AL) is a promising ML paradigm that has the potential to parse through large unlabeled data and help reduce annotation cost in domains where labeling data can be prohibitive. Recently proposed neural network based AL methods use different heuristics to accomplish this goal. In this study, we demonstrate that under identical experimental settings, different types of AL algorithms (uncertainty based, diversity based, and committee based) produce an inconsistent gain over random sampling baseline. Through a variety of experiments, controlling for sources of stochasticity, we show that variance in performance metrics achieved by AL algorithms can lead to results that are not consistent with the previously reported results. We also found that under strong regularization, AL methods show marginal or no advantage over the random sampling baseline under a variety of experimental conditions. Finally, we conclude with a set of recommendations on how to assess the results using a new AL algorithm to ensure results are reproducible and robust under changes in experimental conditions. We share our codes to facilitate AL evaluations. We believe our findings and recommendations will help advance reproducible research in AL using neural networks. We open source our code at https://github.com/PrateekMunjal/TorchAL

Clément Christophe, Praveen K Kanithi, Prateek Munjal et al.

This study presents a comprehensive analysis and comparison of two predominant fine-tuning methodologies - full-parameter fine-tuning and parameter-efficient tuning - within the context of medical Large Language Models (LLMs). We developed and refined a series of LLMs, based on the Llama-2 architecture, specifically designed to enhance medical knowledge retrieval, reasoning, and question-answering capabilities. Our experiments systematically evaluate the effectiveness of these tuning strategies across various well-known medical benchmarks. Notably, our medical LLM Med42 showed an accuracy level of 72% on the US Medical Licensing Examination (USMLE) datasets, setting a new standard in performance for openly available medical LLMs. Through this comparative analysis, we aim to identify the most effective and efficient method for fine-tuning LLMs in the medical domain, thereby contributing significantly to the advancement of AI-driven healthcare applications.

Svetlana Maslenkova, Clement Christophe, Marco AF Pimentel et al.

Large language models offer transformative potential for healthcare, yet their responsible and equitable development depends critically on a deeper understanding of how training data characteristics influence model behavior, including the potential for bias. Current practices in dataset curation and bias assessment often lack the necessary transparency, creating an urgent need for comprehensive evaluation frameworks to foster trust and guide improvements. In this study, we present an in-depth analysis of potential downstream biases in clinical language models, with a focus on differential opioid prescription tendencies across diverse demographic groups, such as ethnicity, gender, and age. As part of this investigation, we introduce HC4: Healthcare Comprehensive Commons Corpus, a novel and extensively curated pretraining dataset exceeding 89 billion tokens. Our evaluation leverages both established general benchmarks and a novel, healthcare-specific methodology, offering crucial insights to support fairness and safety in clinical AI applications.

Davood Karimi, Lana Vasung, Camilo Jaimes et al.

Multi-compartment modeling of diffusion-weighted magnetic resonance imaging measurements is necessary for accurate brain connectivity analysis. Existing methods for estimating the number and orientations of fascicles in an imaging voxel either depend on non-convex optimization techniques that are sensitive to initialization and measurement noise, or are prone to predicting spurious fascicles. In this paper, we propose a machine learning-based technique that can accurately estimate the number and orientations of fascicles in a voxel. Our method can be trained with either simulated or real diffusion-weighted imaging data. Our method estimates the angle to the closest fascicle for each direction in a set of discrete directions uniformly spread on the unit sphere. This information is then processed to extract the number and orientations of fascicles in a voxel. On realistic simulated phantom data with known ground truth, our method predicts the number and orientations of crossing fascicles more accurately than several existing methods. It also leads to more accurate tractography. On real data, our method is better than or compares favorably with standard methods in terms of robustness to measurement down-sampling and also in terms of expert quality assessment of tractography results.

Ahmed H. Shahin, Prateek Munjal, Ling Shao et al.

Semantic segmentation from user inputs has been actively studied to facilitate interactive segmentation for data annotation and other applications. Recent studies have shown that extreme points can be effectively used to encode user inputs. A heat map generated from the extreme points can be appended to the RGB image and input to the model for training. In this study, we present FAIRS -- a new approach to generate object segmentation from user inputs in the form of extreme points and corrective clicks. We propose a novel approach for effectively encoding the user input from extreme points and corrective clicks, in a novel and scalable manner that allows the network to work with a variable number of clicks, including corrective clicks for output refinement. We also integrate a dual attention module with our approach to increase the efficacy of the model in preferentially attending to the objects. We demonstrate that these additions help achieve significant improvements over state-of-the-art in dense object segmentation from user inputs, on multiple large-scale datasets. Through experiments, we demonstrate our method's ability to generate high-quality training data as well as its scalability in incorporating extreme points, guiding clicks, and corrective clicks in a principled manner.

Shadab Khan, Ahmed H. Shahin, Javier Villafruela et al.

To automate the process of segmenting an anatomy of interest, we can learn a model from previously annotated data. The learning-based approach uses annotations to train a model that tries to emulate the expert labeling on a new data set. While tremendous progress has been made using such approaches, labeling of medical images remains a time-consuming and expensive task. In this paper, we evaluate the utility of extreme points in learning to segment. Specifically, we propose a novel approach to compute a confidence map from extreme points that quantitatively encodes the priors derived from extreme points. We use the confidence map as a cue to train a deep neural network based on ResNet-101 and PSP module to develop a class-agnostic segmentation model that outperforms state-of-the-art method that employs extreme points as a cue. Further, we evaluate a realistic use-case by using our model to generate training data for supervised learning (U-Net) and observed that U-Net performs comparably when trained with either the generated data or the ground truth data. These findings suggest that models trained using cues can be used to generate reliable training data.

Seyed Sadegh Mohseni Salehi, Shadab Khan, Deniz Erdogmus et al.

With an aim to increase the capture range and accelerate the performance of state-of-the-art inter-subject and subject-to-template 3D registration, we propose deep learning-based methods that are trained to find the 3D position of arbitrarily oriented subjects or anatomy based on slices or volumes of medical images. For this, we propose regression CNNs that learn to predict the angle-axis representation of 3D rotations and translations using image features. We use and compare mean square error and geodesic loss to train regression CNNs for 3D pose estimation used in two different scenarios: slice-to-volume registration and volume-to-volume registration. Our results show that in such registration applications that are amendable to learning, the proposed deep learning methods with geodesic loss minimization can achieve accurate results with a wide capture range in real-time (<100ms). We also tested the generalization capability of the trained CNNs on an expanded age range and on images of newborn subjects with similar and different MR image contrasts. We trained our models on T2-weighted fetal brain MRI scans and used them to predict the 3D pose of newborn brains based on T1-weighted MRI scans. We showed that the trained models generalized well for the new domain when we performed image contrast transfer through a conditional generative adversarial network. This indicates that the domain of application of the trained deep regression CNNs can be further expanded to image modalities and contrasts other than those used in training. A combination of our proposed methods with accelerated optimization-based registration algorithms can dramatically enhance the performance of automatic imaging devices and image processing methods of the future.