David Clifton

Omid Rohanian, Hannah Jauncey, Mohammadmahdi Nouriborji et al.

Processing information locked within clinical health records is a challenging task that remains an active area of research in biomedical NLP. In this work, we evaluate a broad set of machine learning techniques ranging from simple RNNs to specialised transformers such as BioBERT on a dataset containing clinical notes along with a set of annotations indicating whether a sample is cancer-related or not. Furthermore, we specifically employ efficient fine-tuning methods from NLP, namely, bottleneck adapters and prompt tuning, to adapt the models to our specialised task. Our evaluations suggest that fine-tuning a frozen BERT model pre-trained on natural language and with bottleneck adapters outperforms all other strategies, including full fine-tuning of the specialised BioBERT model. Based on our findings, we suggest that using bottleneck adapters in low-resource situations with limited access to labelled data or processing capacity could be a viable strategy in biomedical text mining. The code used in the experiments are going to be made available at https://github.com/omidrohanian/bottleneck-adapters.

Yining Hua, Hongbin Na, Zehan Li et al.

Large language models (LLMs) are emerging as promising tools for mental health care, offering scalable support through their ability to generate human-like responses. However, the effectiveness of these models in clinical settings remains unclear. This scoping review aimed to assess the current generative applications of LLMs in mental health care, focusing on studies where these models were tested with human participants in real-world scenarios. A systematic search across APA PsycNet, Scopus, PubMed, and Web of Science identified 726 unique articles, of which 17 met the inclusion criteria. These studies encompassed applications such as clinical assistance, counseling, therapy, and emotional support. However, the evaluation methods were often non-standardized, with most studies relying on ad hoc scales that limit comparability and robustness. Privacy, safety, and fairness were also frequently underexplored. Moreover, reliance on proprietary models, such as OpenAI's GPT series, raises concerns about transparency and reproducibility. While LLMs show potential in expanding mental health care access, especially in underserved areas, the current evidence does not fully support their use as standalone interventions. More rigorous, standardized evaluations and ethical oversight are needed to ensure these tools can be safely and effectively integrated into clinical practice.

Sean Wu, Pan Lu, Yupeng Chen et al.

Artificial intelligence (AI) is increasingly embedded in scientific discovery, yet whether it can anticipate scientific progress remains unclear. To study this question, we introduce a temporally grounded evaluation framework for forecasting scientific progress under controlled knowledge constraints. We present CUSP (Cutoff-conditioned Unseen Scientific Progress), a multi-disciplinary and event-level benchmark that evaluates scientific forecasting in AI systems through feasibility assessment, mechanistic reasoning, generative solution design, and temporal prediction. Across 4,760 scientific events, we observe systematic and domain-dependent limitations in current frontier models. While models can identify plausible research directions from competing candidates, they fail to reliably predict whether scientific advances will be realized and systematically misestimate when they will occur. Performance is highly heterogeneous across domains, with the timing of AI progress more predictable than advances in biology, chemistry, and physics. Performance is largely insensitive to whether events occur before or after the training cutoff, suggesting these limitations cannot be explained solely by knowledge exposure in training data. Under controlled information access, additional pre-cutoff knowledge improves performance but does not close the gap to full-information settings, which becomes more pronounced for high-citation advances. Models also exhibit systematic overconfidence and strong response biases, indicating unreliable uncertainty estimation. Taken together, current AI systems fall short as predictive tools for scientific progress. Access to prior knowledge does not translate into reliable forecasting, and performance benefits more from post-event information than from forward-looking prediction.

Mohammadmahdi Nouriborji, Omid Rohanian, David Clifton

This paper outlines the system using which team Nowruz participated in SemEval 2022 Task 7 Identifying Plausible Clarifications of Implicit and Underspecified Phrases for both subtasks A and B. Using a pre-trained transformer as a backbone, the model targeted the task of multi-task classification and ranking in the context of finding the best fillers for a cloze task related to instructional texts on the website Wikihow. The system employed a combination of two ordinal regression components to tackle this task in a multi-task learning scenario. According to the official leaderboard of the shared task, this system was ranked 5th in the ranking and 7th in the classification subtasks out of 21 participating teams. With additional experiments, the models have since been further optimised.

Niall Taylor, Upamanyu Ghose, Omid Rohanian et al.

The entry of large language models (LLMs) into research and commercial spaces has led to a trend of ever-larger models, with initial promises of generalisability, followed by a widespread desire to downsize and create specialised models without the need for complete fine-tuning, using Parameter Efficient Fine-tuning (PEFT) methods. We present an investigation into the suitability of different PEFT methods to clinical decision-making tasks, across a range of model sizes, including extremely small models with as few as $25$ million parameters. Our analysis shows that the performance of most PEFT approaches varies significantly from one task to another, with the exception of LoRA, which maintains relatively high performance across all model sizes and tasks, typically approaching or matching full fine-tuned performance. The effectiveness of PEFT methods in the clinical domain is evident, particularly for specialised models which can operate on low-cost, in-house computing infrastructure. The advantages of these models, in terms of speed and reduced training costs, dramatically outweighs any performance gain from large foundation LLMs. Furthermore, we highlight how domain-specific pre-training interacts with PEFT methods and model size, and discuss how these factors interplay to provide the best efficiency-performance trade-off. Full code available at: tbd.

Ruilin Yao, Bo Zhang, Jirui Huang et al.

Multimodal Large Language Models (MLLMs) have achieved significant advances in integrating visual and linguistic information, yet their ability to reason about complex and real-world scenarios remains limited. The existing benchmarks are usually constructed in the task-oriented manner without guarantee that different task samples come from the same data distribution, thus they often fall short in evaluating the synergistic effects of lower-level perceptual capabilities on higher-order reasoning. To lift this limitation, we contribute Lens, a multi-level benchmark with 3.4K contemporary images and 60K+ human-authored questions covering eight tasks and 12 daily scenarios, forming three progressive task tiers, i.e., perception, understanding, and reasoning. One feature is that each image is equipped with rich annotations for all tasks. Thus, this dataset intrinsically supports to evaluate MLLMs to handle image-invariable prompts, from basic perception to compositional reasoning. In addition, our images are manully collected from the social media, in which 53% were published later than Jan. 2025. We evaluate 15+ frontier MLLMs such as Qwen2.5-VL-72B, InternVL3-78B, GPT-4o and two reasoning models QVQ-72B-preview and Kimi-VL. These models are released later than Dec. 2024, and none of them achieve an accuracy greater than 60% in the reasoning tasks. Project page: https://github.com/Lens4MLLMs/lens. ICCV 2025 workshop page: https://lens4mllms.github.io/mars2-workshop-iccv2025/

Xin Wang, Ling Feng, Huijun Zhang et al.

Stress haunts people in modern society, which may cause severe health issues if left unattended. With social media becoming an integral part of daily life, leveraging social media to detect stress has gained increasing attention. While the majority of the work focuses on classifying stress states and stress categories, this study introduce a new task aimed at estimating more specific stressors (like exam, writing paper, etc.) through users' posts on social media. Unfortunately, the diversity of stressors with many different classes but a few examples per class, combined with the consistent arising of new stressors over time, hinders the machine understanding of stressors. To this end, we cast the stressor estimation problem within a practical scenario few-shot learning setting, and propose a novel meta-learning based stressor estimation framework that is enhanced by a meta-knowledge inheritance mechanism. This model can not only learn generic stressor context through meta-learning, but also has a good generalization ability to estimate new stressors with little labeled data. A fundamental breakthrough in our approach lies in the inclusion of the meta-knowledge inheritance mechanism, which equips our model with the ability to prevent catastrophic forgetting when adapting to new stressors. The experimental results show that our model achieves state-of-the-art performance compared with the baselines. Additionally, we construct a social media-based stressor estimation dataset that can help train artificial intelligence models to facilitate human well-being. The dataset is now public at \href{https://www.kaggle.com/datasets/xinwangcs/stressor-cause-of-mental-health-problem-dataset}{\underline{Kaggle}} and \href{https://huggingface.co/datasets/XinWangcs/Stressor}{\underline{Hugging Face}}.

Pafue Christy Nganjimi, Andrew Soltan, Danielle Belgrave et al.

Dataset condensation constructs compact synthetic datasets that retain the training utility of large real-world datasets, enabling efficient model development and potentially supporting downstream research in governed domains such as healthcare. Trajectory matching (TM) is a widely used condensation approach that supervises synthetic data using changes in model parameters observed during training on real data, yet the structure of this supervision signal remains poorly understood. In this paper, we provide a geometric characterisation of trajectory matching, showing that a fixed synthetic dataset can only reproduce a limited span of such training-induced parameter changes. When the resulting supervision signal is spectrally broad, this creates a conditional representability bottleneck. Motivated by this mismatch, we propose Bezier Trajectory Matching (BTM), which replaces SGD trajectories with quadratic Bezier trajectory surrogates between initial and final model states. These surrogates are optimised to reduce average loss along the path while replacing broad SGD-derived supervision with a more structured, lower-rank signal that is better aligned with the optimisation constraints of a fixed synthetic dataset, and they substantially reduce trajectory storage. Experiments on five clinical datasets demonstrate that BTM consistently matches or improves upon standard trajectory matching, with the largest gains in low-prevalence and low-synthetic-budget settings. These results indicate that effective trajectory matching depends on structuring the supervision signal rather than reproducing stochastic optimisation paths.

Xin Wang, Boyan Gao, Yi Dai et al.

Stress is a pervasive global health issue that can lead to severe mental health problems. Early detection offers timely intervention and prevention of stress-related disorders. The current early detection models perform "black box" inference suffering from limited explainability and trust which blocks the real-world clinical application. Thanks to the generative properties introduced by the Large Language Models (LLMs), the decision and the prediction from such models are semi-interpretable through the corresponding description. However, the existing LLMs are mostly trained for general purposes without the guidance of psychological cognitive theory. To this end, we first highlight the importance of prior theory with the observation of performance boosted by the chain-of-thoughts tailored for stress detection. This method termed Cognition Chain explicates the generation of stress through a step-by-step cognitive perspective based on cognitive appraisal theory with a progress pipeline: Stimulus $\rightarrow$ Evaluation $\rightarrow$ Reaction $\rightarrow$ Stress State, guiding LLMs to provide comprehensive reasoning explanations. We further study the benefits brought by the proposed Cognition Chain format by utilising it as a synthetic dataset generation template for LLMs instruction-tuning and introduce CogInstruct, an instruction-tuning dataset for stress detection. This dataset is developed using a three-stage self-reflective annotation pipeline that enables LLMs to autonomously generate and refine instructional data. By instruction-tuning Llama3 with CogInstruct, we develop CogLLM, an explainable stress detection model. Evaluations demonstrate that CogLLM achieves outstanding performance while enhancing explainability. Our work contributes a novel approach by integrating cognitive theories into LLM reasoning processes, offering a promising direction for future explainable AI research.

James Anibal, Hannah Huth, Ming Li et al.

Artificial intelligence (AI) models trained on audio data may have the potential to rapidly perform clinical tasks, enhancing medical decision-making and potentially improving outcomes through early detection. Existing technologies depend on limited datasets collected with expensive recording equipment in high-income countries, which challenges deployment in resource-constrained, high-volume settings where audio data may have a profound impact on health equity. This report introduces a novel data type and a corresponding collection system that captures health data through guided questions using only a mobile/web application. The app facilitates the collection of an audio electronic health record (Voice EHR) which may contain complex biomarkers of health from conventional voice/respiratory features, speech patterns, and spoken language with semantic meaning and longitudinal context, potentially compensating for the typical limitations of unimodal clinical datasets. This report presents the application used for data collection, initial experiments on data quality, and case studies which demonstrate the potential of voice EHR to advance the scalability/diversity of audio AI.

Edward Phillips, Sean Wu, Soheila Molaei et al.

Large language models demonstrate impressive results across diverse tasks but are still known to hallucinate, generating linguistically plausible but incorrect answers to questions. Uncertainty quantification has been proposed as a strategy for hallucination detection, but no existing black-box approach provides estimates for both global and local uncertainty. The former attributes uncertainty to a batch of responses, while the latter attributes uncertainty to individual responses. Current local methods typically rely on white-box access to internal model states, whilst black-box methods only provide global uncertainty estimates. We introduce a geometric framework to address this, based on archetypal analysis of batches of responses sampled with only black-box model access. At the global level, we propose Geometric Volume, which measures the convex hull volume of archetypes derived from response embeddings. At the local level, we propose Geometric Suspicion, which ranks responses by reliability and enables hallucination reduction through preferential response selection. Unlike prior dispersion methods which yield only a single global score, our approach provides semantic boundary points which have utility for attributing reliability to individual responses. Experiments show that our framework performs comparably to or better than prior methods on short form question-answering datasets, and achieves superior results on medical datasets where hallucinations carry particularly critical risks. We also provide theoretical justification by proving a link between convex hull volume and entropy.

Pancheng Zhao, Deng-Ping Fan, Shupeng Cheng et al.

Deep learning is developing rapidly and handling common computer vision tasks well. It is time to pay attention to more complex vision tasks, as model size, knowledge, and reasoning capabilities continue to improve. In this paper, we introduce and review a family of complex tasks, termed Concealed Dense Prediction (CDP), which has great value in agriculture, industry, etc. CDP's intrinsic trait is that the targets are concealed in their surroundings, thus fully perceiving them requires fine-grained representations, prior knowledge, auxiliary reasoning, etc. The contributions of this review are three-fold: (i) We introduce the scope, characteristics, and challenges specific to CDP tasks and emphasize their essential differences from generic vision tasks. (ii) We develop a taxonomy based on concealment counteracting to summarize deep learning efforts in CDP through experiments on three tasks. We compare 25 state-of-the-art methods across 12 widely used concealed datasets. (iii) We discuss the potential applications of CDP in the large model era and summarize 6 potential research directions. We offer perspectives for the future development of CDP by constructing a large-scale multimodal instruction fine-tuning dataset, CvpINST, and a concealed visual perception agent, CvpAgent.

Pramit Saha, Felix Wagner, Divyanshu Mishra et al.

Effective training of large Vision-Language Models (VLMs) on resource-constrained client devices in Federated Learning (FL) requires the usage of parameter-efficient fine-tuning (PEFT) strategies. To this end, we demonstrate the impact of two factors \textit{viz.}, client-specific layer importance score that selects the most important VLM layers for fine-tuning and inter-client layer diversity score that encourages diverse layer selection across clients for optimal VLM layer selection. We first theoretically motivate and leverage the principal eigenvalue magnitude of layerwise Neural Tangent Kernels and show its effectiveness as client-specific layer importance score. Next, we propose a novel layer updating strategy dubbed F$^3$OCUS that jointly optimizes the layer importance and diversity factors by employing a data-free, multi-objective, meta-heuristic optimization on the server. We explore 5 different meta-heuristic algorithms and compare their effectiveness for selecting model layers and adapter layers towards PEFT-FL. Furthermore, we release a new MedVQA-FL dataset involving overall 707,962 VQA triplets and 9 modality-specific clients and utilize it to train and evaluate our method. Overall, we conduct more than 10,000 client-level experiments on 6 Vision-Language FL task settings involving 58 medical image datasets and 4 different VLM architectures of varying sizes to demonstrate the effectiveness of the proposed method.

Boyan Gao, Xin Wang, Yibo Yang et al.

Large Language Models (LLMs) have demonstrated remarkable performance in real-world applications. However, adapting LLMs to novel tasks via fine-tuning often requires substantial training data and computational resources that are impractical in few-shot scenarios. Existing approaches, such as in-context learning and Parameter-Efficient Fine-Tuning (PEFT), face key limitations: in-context learning introduces additional inference computational overhead with limited performance gains, while PEFT models are prone to overfitting on the few demonstration examples. In this work, we reinterpret the forward pass of LLMs as an optimization process, a sequence of preconditioned gradient descent steps refining internal representations. Based on this connection, we propose Optimization-Inspired Few-Shot Adaptation (OFA), integrating a parameterization that learns preconditioners without introducing additional trainable parameters, and an objective that improves optimization efficiency by learning preconditioners based on a convergence bound, while simultaneously steering the optimization path toward the flat local minimum. Our method overcomes both issues of ICL-based and PEFT-based methods, and demonstrates superior performance over the existing methods on a variety of few-shot adaptation tasks in experiments.

Bradley Segal, Joshua Fieggen, David Clifton et al. · oxford

Ensuring the generalisability of clinical machine learning (ML) models across diverse healthcare settings remains a significant challenge due to variability in patient demographics, disease prevalence, and institutional practices. Existing model evaluation approaches often rely on real-world datasets, which are limited in availability, embed confounding biases, and lack the flexibility needed for systematic experimentation. Furthermore, while generative models aim for statistical realism, they often lack transparency and explicit control over factors driving distributional shifts. In this work, we propose a novel structured synthetic data framework designed for the controlled benchmarking of model robustness, fairness, and generalisability. Unlike approaches focused solely on mimicking observed data, our framework provides explicit control over the data generating process, including site-specific prevalence variations, hierarchical subgroup effects, and structured feature interactions. This enables targeted investigation into how models respond to specific distributional shifts and potential biases. Through controlled experiments, we demonstrate the framework's ability to isolate the impact of site variations, support fairness-aware audits, and reveal generalisation failures, particularly highlighting how model complexity interacts with site-specific effects. This work contributes a reproducible, interpretable, and configurable tool designed to advance the reliable deployment of ML in clinical settings.

Archie Licudi, Anshul Thakur, Soheila Molaei et al.

Asynchronous federated learning (FL) with heterogeneous clients faces two key issues: curvature-induced loss barriers encountered by standard linear parameter interpolation techniques (e.g. FedAvg) and interference from stale updates misaligned with the server's current optimisation state. To alleviate these issues, we introduce a geometric framework that casts aggregation as curve learning in a Riemannian model space and decouples trajectory selection from update conflict resolution. Within this, we propose AsyncBezier, which replaces linear aggregation with low-degree polynomial (Bezier) trajectories to bypass loss barriers, and OrthoDC, which projects delayed updates via inner product-based orthogonality to reduce interference. We establish framework-level convergence guarantees covering each variant given simple assumptions on their components. On three datasets spanning general-purpose and healthcare domains, including LEAF Shakespeare and FEMNIST, our approach consistently improves accuracy and client fairness over strong asynchronous baselines; finally, we show that these gains are preserved even when other methods are allocated a higher local compute budget.

Alex Youssef, Michael Pencina, Anshul Thakur et al.

External validation is often recommended to ensure the generalizability of ML models. However, it neither guarantees generalizability nor equates to a model's clinical usefulness (the ultimate goal of any clinical decision-support tool). External validation is misaligned with current healthcare ML needs. First, patient data changes across time, geography, and facilities. These changes create significant volatility in the performance of a single fixed model (especially for deep learning models, which dominate clinical ML). Second, newer ML techniques, current market forces, and updated regulatory frameworks are enabling frequent updating and monitoring of individual deployed model instances. We submit that external validation is insufficient to establish ML models' safety or utility. Proposals to fix the external validation paradigm do not go far enough. Continued reliance on it as the ultimate test is likely to lead us astray. We propose the MLOps-inspired paradigm of recurring local validation as an alternative that ensures the validity of models while protecting against performance-disruptive data variability. This paradigm relies on site-specific reliability tests before every deployment, followed by regular and recurrent checks throughout the life cycle of the deployed algorithm. Initial and recurrent reliability tests protect against performance-disruptive distribution shifts, and concept drifts that jeopardize patient safety.

Anna Antoniou, Giacomo Dossena, Julia MacMillan et al.

Objective: The use of routinely-acquired medical data for research purposes requires the protection of patient confidentiality via data anonymisation. The objective of this work is to calculate the risk of re-identification arising from a malicious attack to an anonymised dataset, as described below. Methods: We first present an analytical means of estimating the probability of re-identification of a single patient in a k-anonymised dataset of Electronic Health Record (EHR) data. Second, we generalize this solution to obtain the probability of multiple patients being re-identified. We provide synthetic validation via Monte Carlo simulations to illustrate the accuracy of the estimates obtained. Results: The proposed analytical framework for risk estimation provides re-identification probabilities that are in agreement with those provided by simulation in a number of scenarios. Our work is limited by conservative assumptions which inflate the re-identification probability. Discussion: Our estimates show that the re-identification probability increases with the proportion of the dataset maliciously obtained and that it has an inverse relationship with the equivalence class size. Our recursive approach extends the applicability domain to the general case of a multi-patient re-identification attack in an arbitrary k-anonymisation scheme. Conclusion: We prescribe a systematic way to parametrize the k-anonymisation process based on a pre-determined re-identification probability. We observed that the benefits of a reduced re-identification risk that come with increasing k-size may not be worth the reduction in data granularity when one is considering benchmarking the re-identification probability on the size of the portion of the dataset maliciously obtained by the adversary.

Omid Rohanian, Samaneh Kouchaki, Andrew Soltan et al.

Early detection of COVID-19 is an ongoing area of research that can help with triage, monitoring and general health assessment of potential patients and may reduce operational strain on hospitals that cope with the coronavirus pandemic. Different machine learning techniques have been used in the literature to detect coronavirus using routine clinical data (blood tests, and vital signs). Data breaches and information leakage when using these models can bring reputational damage and cause legal issues for hospitals. In spite of this, protecting healthcare models against leakage of potentially sensitive information is an understudied research area. In this work, we examine two machine learning approaches, intended to predict a patient's COVID-19 status using routinely collected and readily available clinical data. We employ adversarial training to explore robust deep learning architectures that protect attributes related to demographic information about the patients. The two models we examine in this work are intended to preserve sensitive information against adversarial attacks and information leakage. In a series of experiments using datasets from the Oxford University Hospitals, Bedfordshire Hospitals NHS Foundation Trust, University Hospitals Birmingham NHS Foundation Trust, and Portsmouth Hospitals University NHS Trust we train and test two neural networks that predict PCR test results using information from basic laboratory blood tests, and vital signs performed on a patients' arrival to hospital. We assess the level of privacy each one of the models can provide and show the efficacy and robustness of our proposed architectures against a comparable baseline. One of our main contributions is that we specifically target the development of effective COVID-19 detection models with built-in mechanisms in order to selectively protect sensitive attributes against adversarial attacks.

Dani Kiyasseh, Tingting Zhu, David Clifton

Cardiac signals, such as the electrocardiogram, convey a significant amount of information about the health status of a patient which is typically summarized by a clinician in the form of a clinical report, a cumbersome process that is prone to errors. To streamline this routine process, we propose a deep neural network capable of captioning cardiac signals; it receives a cardiac signal as input and generates a clinical report as output. We extend this further to generate multilingual reports. To that end, we create and make publicly available a multilingual clinical report dataset. In the absence of sufficient labelled data, deep neural networks can benefit from a warm-start, or pre-training, procedure in which parameters are first learned in an arbitrary task. We propose such a task in the form of discriminative multilingual pre-training where tokens from clinical reports are randomly replaced with those from other languages and the network is tasked with predicting the language of all tokens. We show that our method performs on par with state-of-the-art pre-training methods such as MLM, ELECTRA, and MARGE, while simultaneously generating diverse and plausible clinical reports. We also demonstrate that multilingual models can outperform their monolingual counterparts, informally terming this beneficial phenomenon as the blessing of multilinguality.

Rasheed el-Bouri, David Eyre, Peter Watkinson et al.

Accurate and reliable prediction of hospital admission location is important due to resource-constraints and space availability in a clinical setting, particularly when dealing with patients who come from the emergency department. In this work we propose a student-teacher network via reinforcement learning to deal with this specific problem. A representation of the weights of the student network is treated as the state and is fed as an input to the teacher network. The teacher network's action is to select the most appropriate batch of data to train the student network on from a training set sorted according to entropy. By validating on three datasets, not only do we show that our approach outperforms state-of-the-art methods on tabular data and performs competitively on image recognition, but also that novel curricula are learned by the teacher network. We demonstrate experimentally that the teacher network can actively learn about the student network and guide it to achieve better performance than if trained alone.