Large reasoning models have recently made significant strides in mathematical and code reasoning, yet their success has not transferred smoothly to the medical domain. While multiple factors contribute to this disparity, a critical issue is the inadequate focus on the quality of intermediate reflection steps, which is particularly crucial in high-stakes medical scenarios. To address this challenge, we propose Med-REFL, a \underline{\textbf{Med}}ical \underline{\textbf{R}}easoning \underline{\textbf{E}}nhancement via self-corrected \underline{\textbf{F}}ine-grained ref\underline{\textbf{L}}ection. Our method leverages a tree-of-thought approach to decompose medical questions into fine-grained reasoning paths, quantitatively evaluating each step and its subsequent reflections. These assessments enable automatic construction of direct preference optimization data, reducing reliance on expensive expert annotations while guiding models to identify and correct reasoning errors. Experimental results on the MedQA-USMLE benchmark demonstrate Med-REFL achieves consistent improvements, with average gains up to 4.11\%. Notably, it further boosts the state-of-the-art performance of 7B/8B models by an additional 4.13\%. Furthermore, Med-REFL exhibits strong generalization capabilities and robustness across several challenging medical question-answering datasets. Our work illustrates that prioritizing reflection quality leads to more accurate and trustworthy reasoning in medical AI applications. Checkpoints, code, and data can be found in https://github.com/TianYin123/Med-REFL.
Though reasoning-based large language models (LLMs) have excelled in mathematics and programming, their capabilities in knowledge-intensive medical question answering remain underexplored. To address this, we introduce ReasonMed, the largest medical reasoning dataset, comprising 370k high-quality examples distilled from 1.7 million initial reasoning paths generated by various LLMs. ReasonMed is constructed through a \textit{multi-agent verification and refinement process}, where we design an \textit{Error Refiner} to enhance the reasoning paths by identifying and correcting error-prone steps flagged by a verifier. Leveraging ReasonMed, we systematically investigate best practices for training medical reasoning models and find that combining detailed Chain-of-Thought (CoT) reasoning with concise answer summaries yields the most effective fine-tuning strategy. Based on this strategy, we train ReasonMed-7B, which sets a new benchmark for sub-10B models, outperforming the prior best by 4.17\% and even exceeding LLaMA3.1-70B on PubMedQA by 4.60\%.
Large language models (LLMs) are reaching expert-level accuracy on medical diagnosis questions, yet their mistakes and the biases behind them pose life-critical risks. Bias linked to race, sex, and socioeconomic status is already well known, but a consistent and automatic testbed for measuring it is missing. To fill this gap, this paper presents AMQA -- an Adversarial Medical Question-Answering dataset -- built for automated, large-scale bias evaluation of LLMs in medical QA. AMQA includes 4,806 medical QA pairs sourced from the United States Medical Licensing Examination (USMLE) dataset, generated using a multi-agent framework to create diverse adversarial descriptions and question pairs. Using AMQA, we benchmark five representative LLMs and find surprisingly substantial disparities: even GPT-4.1, the least biased model tested, answers privileged-group questions over 10 percentage points more accurately than unprivileged ones. Compared with the existing benchmark CPV, AMQA reveals 15% larger accuracy gaps on average between privileged and unprivileged groups. Our dataset and code are publicly available at https://github.com/XY-Showing/AMQA to support reproducible research and advance trustworthy, bias-aware medical AI.
The rapid advancement of language models has demonstrated the potential of artificial intelligence in the healthcare industry. However, small language models struggle with specialized domains in low-resource languages like Persian. While numerous medical-domain websites exist in Persian, no curated dataset or corpus has been available making ours the first of its kind. This study explores the enhancement of medical knowledge in a small language model by leveraging accessible online data, including a crawled corpus from medical magazines and a dataset of real doctor-patient QA pairs. We fine-tuned a baseline model using our curated data to improve its medical knowledge. Benchmark evaluations demonstrate that the fine-tuned model achieves improved accuracy in medical question answering and provides better responses compared to its baseline. This work highlights the potential of leveraging open-access online data to enrich small language models in medical fields, providing a novel solution for Persian medical AI applications suitable for resource-constrained environments.
Medical question answering (QA) is a reasoning-intensive task that remains challenging for large language models (LLMs) due to hallucinations and outdated domain knowledge. Retrieval-Augmented Generation (RAG) provides a promising post-training solution by leveraging external knowledge. However, existing medical RAG systems suffer from two key limitations: (1) a lack of modeling for human-like reasoning behaviors during information retrieval, and (2) reliance on suboptimal medical corpora, which often results in the retrieval of irrelevant or noisy snippets. To overcome these challenges, we propose Discuss-RAG, a plug-and-play module designed to enhance the medical QA RAG system through collaborative agent-based reasoning. Our method introduces a summarizer agent that orchestrates a team of medical experts to emulate multi-turn brainstorming, thereby improving the relevance of retrieved content. Additionally, a decision-making agent evaluates the retrieved snippets before their final integration. Experimental results on four benchmark medical QA datasets show that Discuss-RAG consistently outperforms MedRAG, especially significantly improving answer accuracy by up to 16.67% on BioASQ and 12.20% on PubMedQA. The code is available at: https://github.com/LLM-VLM-GSL/Discuss-RAG.
Large language models (LLMs) are capable of generating plausible explanations of how they arrived at an answer to a question. However, these explanations can misrepresent the model's "reasoning" process, i.e., they can be unfaithful. This, in turn, can lead to over-trust and misuse. We introduce a new approach for measuring the faithfulness of LLM explanations. First, we provide a rigorous definition of faithfulness. Since LLM explanations mimic human explanations, they often reference high-level concepts in the input question that purportedly influenced the model. We define faithfulness in terms of the difference between the set of concepts that LLM explanations imply are influential and the set that truly are. Second, we present a novel method for estimating faithfulness that is based on: (1) using an auxiliary LLM to modify the values of concepts within model inputs to create realistic counterfactuals, and (2) using a Bayesian hierarchical model to quantify the causal effects of concepts at both the example- and dataset-level. Our experiments show that our method can be used to quantify and discover interpretable patterns of unfaithfulness. On a social bias task, we uncover cases where LLM explanations hide the influence of social bias. On a medical question answering task, we uncover cases where LLM explanations provide misleading claims about which pieces of evidence influenced the model's decisions.
Large Language Models (LLMs) have shown impressive performance on existing medical question-answering benchmarks. This high performance makes it increasingly difficult to meaningfully evaluate and differentiate advanced methods. We present MedAgentsBench, a benchmark that focuses on challenging medical questions requiring multi-step clinical reasoning, diagnosis formulation, and treatment planning-scenarios where current models still struggle despite their strong performance on standard tests. Drawing from seven established medical datasets, our benchmark addresses three key limitations in existing evaluations: (1) the prevalence of straightforward questions where even base models achieve high performance, (2) inconsistent sampling and evaluation protocols across studies, and (3) lack of systematic analysis of the interplay between performance, cost, and inference time. Through experiments with various base models and reasoning methods, we demonstrate that the latest thinking models, DeepSeek R1 and OpenAI o3, exhibit exceptional performance in complex medical reasoning tasks. Additionally, advanced search-based agent methods offer promising performance-to-cost ratios compared to traditional approaches. Our analysis reveals substantial performance gaps between model families on complex questions and identifies optimal model selections for different computational constraints. Our benchmark and evaluation framework are publicly available at https://github.com/gersteinlab/medagents-benchmark.
Medication recommendation systems have garnered attention within healthcare for their potential to deliver personalized and efficacious drug combinations based on patient's clinical data. However, existing methodologies encounter challenges in adapting to diverse Electronic Health Records (EHR) systems and effectively utilizing unstructured data, resulting in limited generalization capabilities and suboptimal performance. Recently, interest is growing in harnessing Large Language Models (LLMs) in the medical domain to support healthcare professionals and enhance patient care. Despite the emergence of medical LLMs and their promising results in tasks like medical question answering, their practical applicability in clinical settings, particularly in medication recommendation, often remains underexplored. In this study, we evaluate both general-purpose and medical-specific LLMs for medication recommendation tasks. Our findings reveal that LLMs frequently encounter the challenge of overprescribing, leading to heightened clinical risks and diminished medication recommendation accuracy. To address this issue, we propose Language-Assisted Medication Recommendation (LAMO), which employs a parameter-efficient fine-tuning approach to tailor open-source LLMs for optimal performance in medication recommendation scenarios. LAMO leverages the wealth of clinical information within clinical notes, a resource often underutilized in traditional methodologies. As a result of our approach, LAMO outperforms previous state-of-the-art methods by over 10% in internal validation accuracy. Furthermore, temporal and external validations demonstrate LAMO's robust generalization capabilities across various temporal and hospital contexts. Additionally, an out-of-distribution medication recommendation experiment demonstrates LAMO's remarkable accuracy even with medications outside the training data.
Federated learning (FL) is a popular paradigm for collaborative training which avoids direct data exposure between clients. However, data privacy issues still remain: FL-trained large language models are capable of memorizing and completing phrases and sentences contained in training data when given with their prefixes. Thus, it is possible for adversarial and honest-but-curious clients to recover training data of other participants simply through targeted prompting. In this work, we demonstrate that a popular and simple fine-tuning strategy, low-rank adaptation (LoRA), reduces memorization during FL up to a factor of 10. We study this effect by performing a medical question-answering fine-tuning task and injecting multiple replicas of out-of-distribution sensitive sequences drawn from an external clinical dataset. We observe a reduction in memorization for a wide variety of Llama 2 and 3 models, and find that LoRA can reduce memorization in centralized learning as well. Furthermore, we show that LoRA can be combined with other privacy-preserving techniques such as gradient clipping and Gaussian noising, secure aggregation, and Goldfish loss to further improve record-level privacy while maintaining performance.
In knowledge-intensive tasks, especially in high-stakes domains like medicine and law, it is critical not only to retrieve relevant information but also to provide causal reasoning and explainability. Large language models (LLMs) have achieved remarkable performance in natural language understanding and generation tasks. However, they often suffer from limitations such as difficulty in incorporating new knowledge, generating hallucinations, and explaining their reasoning process. To address these challenges, integrating knowledge graphs with Graph Retrieval-Augmented Generation (Graph RAG) has emerged as an effective solution. Traditional Graph RAG methods often rely on simple graph traversal or semantic similarity, which do not capture causal relationships or align well with the model's internal reasoning steps. This paper proposes a novel pipeline that filters large knowledge graphs to emphasize cause-effect edges, aligns the retrieval process with the model's chain-of-thought (CoT), and enhances reasoning through multi-stage path improvements. Experiments on medical question-answering tasks show consistent gains, with up to a 10\% absolute improvement across multiple large language models (LLMs). This approach demonstrates the value of combining causal reasoning with stepwise retrieval, leading to more interpretable and logically grounded solutions for complex queries.
Several recent works seek to develop foundation models specifically for medical applications, adapting general-purpose large language models (LLMs) and vision-language models (VLMs) via continued pretraining on publicly available biomedical corpora. These works typically claim that such domain-adaptive pretraining (DAPT) improves performance on downstream medical tasks, such as answering medical licensing exam questions. In this paper, we compare ten public "medical" LLMs and two VLMs against their corresponding base models, arriving at a different conclusion: all medical VLMs and nearly all medical LLMs fail to consistently improve over their base models in the zero-/few-shot prompting and supervised fine-tuning regimes for medical question-answering (QA). For instance, across all tasks and model pairs we consider in the 3-shot setting, medical LLMs only outperform their base models in 22.7% of cases, reach a (statistical) tie in 36.8% of cases, and are significantly worse than their base models in the remaining 40.5% of cases. Our conclusions are based on (i) comparing each medical model head-to-head, directly against the corresponding base model; (ii) optimizing the prompts for each model separately in zero-/few-shot prompting; and (iii) accounting for statistical uncertainty in comparisons. While these basic practices are not consistently adopted in the literature, our ablations show that they substantially impact conclusions. Meanwhile, we find that after fine-tuning on specific QA tasks, medical LLMs can show performance improvements, but the benefits do not carry over to tasks based on clinical notes. Our findings suggest that state-of-the-art general-domain models may already exhibit strong medical knowledge and reasoning capabilities, and offer recommendations to strengthen the conclusions of future studies.
Several recent works seek to develop foundation models specifically for medical applications, adapting general-purpose large language models (LLMs) and vision-language models (VLMs) via continued pretraining on publicly available biomedical corpora. These works typically claim that such domain-adaptive pretraining (DAPT) improves performance on downstream medical tasks, such as answering medical licensing exam questions. In this paper, we compare seven public "medical" LLMs and two VLMs against their corresponding base models, arriving at a different conclusion: all medical VLMs and nearly all medical LLMs fail to consistently improve over their base models in the zero-/few-shot prompting regime for medical question-answering (QA) tasks. For instance, across the tasks and model pairs we consider in the 3-shot setting, medical LLMs only outperform their base models in 12.1% of cases, reach a (statistical) tie in 49.8% of cases, and are significantly worse than their base models in the remaining 38.2% of cases. Our conclusions are based on (i) comparing each medical model head-to-head, directly against the corresponding base model; (ii) optimizing the prompts for each model separately; and (iii) accounting for statistical uncertainty in comparisons. While these basic practices are not consistently adopted in the literature, our ablations show that they substantially impact conclusions. Our findings suggest that state-of-the-art general-domain models may already exhibit strong medical knowledge and reasoning capabilities, and offer recommendations to strengthen the conclusions of future studies.
Large language models (LLM) hold significant potential for applications in biomedicine, but they struggle with hallucinations and outdated knowledge. While retrieval-augmented generation (RAG) is generally employed to address these issues, it also has its own set of challenges: (1) LLMs are vulnerable to irrelevant or incorrect context, (2) medical queries are often not well-targeted for helpful information, and (3) retrievers are prone to bias toward the specific source corpus they were trained on. In this study, we present RAG$^2$ (RAtionale-Guided RAG), a new framework for enhancing the reliability of RAG in biomedical contexts. RAG$^2$ incorporates three key innovations: a small filtering model trained on perplexity-based labels of rationales, which selectively augments informative snippets of documents while filtering out distractors; LLM-generated rationales as queries to improve the utility of retrieved snippets; a structure designed to retrieve snippets evenly from a comprehensive set of four biomedical corpora, effectively mitigating retriever bias. Our experiments demonstrate that RAG$^2$ improves the state-of-the-art LLMs of varying sizes, with improvements of up to 6.1\%, and it outperforms the previous best medical RAG model by up to 5.6\% across three medical question-answering benchmarks. Our code is available at https://github.com/dmis-lab/RAG2.
This study explores the potential of using training dynamics as an automated alternative to human annotation for evaluating the quality of training data. The framework used is Data Maps, which classifies data points into categories such as easy-to-learn, hard-to-learn, and ambiguous (Swayamdipta et al., 2020). Swayamdipta et al. (2020) highlight that difficult-to-learn examples often contain errors, and ambiguous cases significantly impact model training. To confirm the reliability of these findings, we replicated the experiments using a challenging dataset, with a focus on medical question answering. In addition to text comprehension, this field requires the acquisition of detailed medical knowledge, which further complicates the task. A comprehensive evaluation was conducted to assess the feasibility and transferability of the Data Maps framework to the medical domain. The evaluation indicates that the framework is unsuitable for addressing datasets' unique challenges in answering medical questions.
In Medical question-answering (QA) tasks, the need for effective systems is pivotal in delivering accurate responses to intricate medical queries. However, existing approaches often struggle to grasp the intricate logical structures and relationships inherent in medical contexts, thus limiting their capacity to furnish precise and nuanced answers. In this work, we address this gap by proposing a novel Abstractive QA system MedLogic-AQA that harnesses First Order Logic (FOL) based rules extracted from both context and questions to generate well-grounded answers. Through initial experimentation, we identified six pertinent first-order logical rules, which were then used to train a Logic-Understanding (LU) model capable of generating logical triples for a given context, question, and answer. These logic triples are then integrated into the training of MedLogic-AQA, enabling effective and coherent reasoning during answer generation. This distinctive fusion of logical reasoning with abstractive QA equips our system to produce answers that are logically sound, relevant, and engaging. Evaluation with respect to both automated and human-based demonstrates the robustness of MedLogic-AQA against strong baselines. Through empirical assessments and case studies, we validate the efficacy of MedLogic-AQA in elevating the quality and comprehensiveness of answers in terms of reasoning as well as informativeness
Explaining Artificial Intelligence (AI) decisions is a major challenge nowadays in AI, in particular when applied to sensitive scenarios like medicine and law. However, the need to explain the rationale behind decisions is a main issue also for human-based deliberation as it is important to justify \textit{why} a certain decision has been taken. Resident medical doctors for instance are required not only to provide a (possibly correct) diagnosis, but also to explain how they reached a certain conclusion. Developing new tools to aid residents to train their explanation skills is therefore a central objective of AI in education. In this paper, we follow this direction, and we present, to the best of our knowledge, the first multilingual dataset for Medical Question Answering where correct and incorrect diagnoses for a clinical case are enriched with a natural language explanation written by doctors. These explanations have been manually annotated with argument components (i.e., premise, claim) and argument relations (i.e., attack, support), resulting in the Multilingual CasiMedicos-Arg dataset which consists of 558 clinical cases in four languages (English, Spanish, French, Italian) with explanations, where we annotated 5021 claims, 2313 premises, 2431 support relations, and 1106 attack relations. We conclude by showing how competitive baselines perform over this challenging dataset for the argument mining task.
The recent success of large language models (LLMs) and the scaling law has led to a widespread adoption of larger models. Particularly in the healthcare industry, there is an increasing demand for locally operated LLMs due to security concerns. However, the majority of high quality open-source LLMs have a size of 70B parameters, imposing significant financial burdens on users for GPU preparation and operation. To overcome these issues, we present a medical adaptation based on the recent 7B models, which enables the operation in low computational resources. We compare the performance on medical question-answering benchmarks in two languages (Japanese and English), demonstrating that its scores reach parity with or surpass those of currently existing medical LLMs that are ten times larger. We find that fine-tuning an English-centric base model on Japanese medical dataset improves the score in both language, supporting the effect of cross-lingual knowledge transfer. We hope that this study will alleviate financial challenges, serving as a stepping stone for clinical institutions to practically utilize LLMs locally. Our evaluation code is available at https://github.com/stardust-coder/japanese-lm-med-harness.
Large language models (LLMs) have significantly advanced natural language processing tasks, yet they are susceptible to generating inaccurate or unreliable responses, a phenomenon known as hallucination. In critical domains such as health and medicine, these hallucinations can pose serious risks. This paper introduces HALO, a novel framework designed to enhance the accuracy and reliability of medical question-answering (QA) systems by focusing on the detection and mitigation of hallucinations. Our approach generates multiple variations of a given query using LLMs and retrieves relevant information from external open knowledge bases to enrich the context. We utilize maximum marginal relevance scoring to prioritize the retrieved context, which is then provided to LLMs for answer generation, thereby reducing the risk of hallucinations. The integration of LangChain further streamlines this process, resulting in a notable and robust increase in the accuracy of both open-source and commercial LLMs, such as Llama-3.1 (from 44% to 65%) and ChatGPT (from 56% to 70%). This framework underscores the critical importance of addressing hallucinations in medical QA systems, ultimately improving clinical decision-making and patient care. The open-source HALO is available at: https://github.com/ResponsibleAILab/HALO.
Black-box large language models (LLMs) are increasingly deployed in various environments, making it essential for these models to effectively convey their confidence and uncertainty, especially in high-stakes settings. However, these models often exhibit overconfidence, leading to potential risks and misjudgments. Existing techniques for eliciting and calibrating LLM confidence have primarily focused on general reasoning datasets, yielding only modest improvements. Accurate calibration is crucial for informed decision-making and preventing adverse outcomes but remains challenging due to the complexity and variability of tasks these models perform. In this work, we investigate the miscalibration behavior of black-box LLMs within the healthcare setting. We propose a novel method, \textit{Atypical Presentations Recalibration}, which leverages atypical presentations to adjust the model's confidence estimates. Our approach significantly improves calibration, reducing calibration errors by approximately 60\% on three medical question answering datasets and outperforming existing methods such as vanilla verbalized confidence, CoT verbalized confidence and others. Additionally, we provide an in-depth analysis of the role of atypicality within the recalibration framework.
Fine-tuning Large Language Models (LLMs) incurs considerable training costs, driving the need for data-efficient training with optimised data ordering. Human-inspired strategies offer a solution by organising data based on human learning practices. This study evaluates the fine-tuning efficiency of five human-inspired strategies across four language models, three datasets, and both human- and LLM-labelled data in the context of medical question answering. These strategies achieve the best accuracy gain of 1.81% and an average gain of 1.02% across datasets, with interleaved strategies delivering the best average results. However, the best strategy varies across model-dataset combinations, limiting the generalisability of the effects of any single strategy. Additionally, LLM-defined question difficulty outperforms human-defined labels in curriculum-based learning, showing the potential of model-generated data as a cost-effective alternative for optimising fine-tuning.