Viktor Schlegel

Viktor Schlegel, Hao Li, Yuping Wu et al. · tencent-ai

This paper describes PULSAR, our system submission at the ImageClef 2023 MediQA-Sum task on summarising patient-doctor dialogues into clinical records. The proposed framework relies on domain-specific pre-training, to produce a specialised language model which is trained on task-specific natural data augmented by synthetic data generated by a black-box LLM. We find limited evidence towards the efficacy of domain-specific pre-training and data augmentation, while scaling up the language model yields the best performance gains. Our approach was ranked second and third among 13 submissions on task B of the challenge. Our code is available at https://github.com/yuping-wu/PULSAR.

Thanh-Tung Nguyen, Viktor Schlegel, Abhinav Kashyap et al.

Clinical notes are assigned ICD codes - sets of codes for diagnoses and procedures. In the recent years, predictive machine learning models have been built for automatic ICD coding. However, there is a lack of widely accepted benchmarks for automated ICD coding models based on large-scale public EHR data. This paper proposes a public benchmark suite for ICD-10 coding using a large EHR dataset derived from MIMIC-IV, the most recent public EHR dataset. We implement and compare several popular methods for ICD coding prediction tasks to standardize data preprocessing and establish a comprehensive ICD coding benchmark dataset. This approach fosters reproducibility and model comparison, accelerating progress toward employing automated ICD coding in future studies. Furthermore, we create a new ICD-9 benchmark using MIMIC-IV data, providing more data points and a higher number of ICD codes than MIMIC-III. Our open-source code offers easy access to data processing steps, benchmark creation, and experiment replication for those with MIMIC-IV access, providing insights, guidance, and protocols to efficiently develop ICD coding models.

Hao Li, Yuping Wu, Viktor Schlegel et al. · tencent-ai

Medical progress notes play a crucial role in documenting a patient's hospital journey, including his or her condition, treatment plan, and any updates for healthcare providers. Automatic summarisation of a patient's problems in the form of a problem list can aid stakeholders in understanding a patient's condition, reducing workload and cognitive bias. BioNLP 2023 Shared Task 1A focuses on generating a list of diagnoses and problems from the provider's progress notes during hospitalisation. In this paper, we introduce our proposed approach to this task, which integrates two complementary components. One component employs large language models (LLMs) for data augmentation; the other is an abstractive summarisation LLM with a novel pre-training objective for generating the patients' problems summarised as a list. Our approach was ranked second among all submissions to the shared task. The performance of our model on the development and test datasets shows that our approach is more robust on unknown data, with an improvement of up to 3.1 points over the same size of the larger model.

Federico Tavella, Viktor Schlegel, Marta Romeo et al.

Signed Language Processing (SLP) concerns the automated processing of signed languages, the main means of communication of Deaf and hearing impaired individuals. SLP features many different tasks, ranging from sign recognition to translation and production of signed speech, but has been overlooked by the NLP community thus far. In this paper, we bring to attention the task of modelling the phonology of sign languages. We leverage existing resources to construct a large-scale dataset of American Sign Language signs annotated with six different phonological properties. We then conduct an extensive empirical study to investigate whether data-driven end-to-end and feature-based approaches can be optimised to automatically recognise these properties. We find that, despite the inherent challenges of the task, graph-based neural networks that operate over skeleton features extracted from raw videos are able to succeed at the task to a varying degree. Most importantly, we show that this performance pertains even on signs unobserved during training.

Aishik Nagar, Yutong Liu, Andy T. Liu et al.

Medical abstractive summarization faces the challenge of balancing faithfulness and informativeness. Current methods often sacrifice key information for faithfulness or introduce confabulations when prioritizing informativeness. While recent advancements in techniques like in-context learning (ICL) and fine-tuning have improved medical summarization, they often overlook crucial aspects such as faithfulness and informativeness without considering advanced methods like model reasoning and self-improvement. Moreover, the field lacks a unified benchmark, hindering systematic evaluation due to varied metrics and datasets. This paper addresses these gaps by presenting a comprehensive benchmark of six advanced abstractive summarization methods across three diverse datasets using five standardized metrics. Building on these findings, we propose uMedSum, a modular hybrid summarization framework that introduces novel approaches for sequential confabulation removal followed by key missing information addition, ensuring both faithfulness and informativeness. Our work improves upon previous GPT-4-based state-of-the-art (SOTA) medical summarization methods, significantly outperforming them in both quantitative metrics and qualitative domain expert evaluations. Notably, we achieve an average relative performance improvement of 11.8% in reference-free metrics over the previous SOTA. Doctors prefer uMedSum's summaries 6 times more than previous SOTA in difficult cases where there are chances of confabulations or missing information. These results highlight uMedSum's effectiveness and generalizability across various datasets and metrics, marking a significant advancement in medical summarization.

Youngmok Ha, Viktor Schlegel, Yidan Sun et al.

While Local Differential Privacy (LDP) serves as a foundational primitive for distributed data collection, its stringent noise injection requirement often leads to severe degradation in data utility. This degradation stems from the task-agnostic nature of conventional LDP mechanisms, which inject noise uniformly across all dimensions regardless of their relative importance to the downstream objective. To address this issue, we propose a novel approach that mitigates noise in task-relevant subspaces of the data representation. Our method identifies task-critical subspaces via the Jacobian matrix of the public downstream model, selectively attenuates noise along those dimensions, and reshapes the isotropic noise of standard LDP into an anisotropic distribution. This method preserves the uniform per-dimension privacy budget while heterogeneously modulating noise impact across dimensions, thereby substantially enhancing data utility. Furthermore, our approach generalizes to both linear and non-linear models and integrates seamlessly with existing mechanisms. Extensive experiments on CIFAR-10-C (Brightness corruption at the highest severity level 5) demonstrate that integrating our approach improves the utility of PrivUnit2 and PrivUnitG by approximately 20\% at $ε=7.5$. The source code is available at \url{https://github.com/ymha/jacobian-anr-ldp}.

Erick Mendez Guzman, Viktor Schlegel, Riza Batista-Navarro

Forced labour is the most common type of modern slavery, and it is increasingly gaining the attention of the research and social community. Recent studies suggest that artificial intelligence (AI) holds immense potential for augmenting anti-slavery action. However, AI tools need to be developed transparently in cooperation with different stakeholders. Such tools are contingent on the availability and access to domain-specific data, which are scarce due to the near-invisible nature of forced labour. To the best of our knowledge, this paper presents the first openly accessible English corpus annotated for multi-class and multi-label forced labour detection. The corpus consists of 989 news articles retrieved from specialised data sources and annotated according to risk indicators defined by the International Labour Organization (ILO). Each news article was annotated for two aspects: (1) indicators of forced labour as classification labels and (2) snippets of the text that justify labelling decisions. We hope that our data set can help promote research on explainability for multi-class and multi-label text classification. In this work, we explain our process for collecting the data underpinning the proposed corpus, describe our annotation guidelines and present some statistical analysis of its content. Finally, we summarise the results of baseline experiments based on different variants of the Bidirectional Encoder Representation from Transformer (BERT) model.

Viktor Schlegel, Kamen V. Pavlov, Ian Pratt-Hartmann

State-of-the-art deep-learning-based approaches to Natural Language Processing (NLP) are credited with various capabilities that involve reasoning with natural language texts. In this paper we carry out a large-scale empirical study investigating the detection of formally valid inferences in controlled fragments of natural language for which the satisfiability problem becomes increasingly complex. We find that, while transformer-based language models perform surprisingly well in these scenarios, a deeper analysis re-veals that they appear to overfit to superficial patterns in the data rather than acquiring the logical principles governing the reasoning in these fragments.

Viktor Schlegel, Erick Mendez-Guzman, Riza Batista-Navarro

Progress on many Natural Language Processing (NLP) tasks, such as text classification, is driven by objective, reproducible and scalable evaluation via publicly available benchmarks. However, these are not always representative of real-world scenarios where text classifiers are employed, such as sentiment analysis or misinformation detection. In this position paper, we put forward two points that aim to alleviate this problem. First, we propose to extend text classification benchmarks to evaluate the explainability of text classifiers. We review challenges associated with objectively evaluating the capabilities to produce valid explanations which leads us to the second main point: We propose to ground these benchmarks in human-centred applications, for example by using social media, gamification or to learn explainability metrics from human judgements.

Aishik Nagar, Viktor Schlegel, Thanh-Tung Nguyen et al.

Large Language Models (LLMs) are increasingly adopted for applications in healthcare, reaching the performance of domain experts on tasks such as question answering and document summarisation. Despite their success on these tasks, it is unclear how well LLMs perform on tasks that are traditionally pursued in the biomedical domain, such as structured information extraction. To bridge this gap, in this paper, we systematically benchmark LLM performance in Medical Classification and Named Entity Recognition (NER) tasks. We aim to disentangle the contribution of different factors to the performance, particularly the impact of LLMs' task knowledge and reasoning capabilities, their (parametric) domain knowledge, and addition of external knowledge. To this end, we evaluate various open LLMs - including BioMistral and Llama-2 models - on a diverse set of biomedical datasets, using standard prompting, Chain of-Thought (CoT) and Self Consistency based reasoning as well as Retrieval-Augmented Generation (RAG) with PubMed and Wikipedia corpora. Counter intuitively, our results reveal that standard prompting consistently outperforms more complex techniques across both tasks, laying bare the limitations in the current application of CoT, self-consistency and RAG in the biomedical domain. Our findings suggest that advanced prompting methods developed for knowledge- or reasoning-intensive tasks, such as CoT or RAG, are not easily portable to biomedical tasks where precise structured outputs are required. This highlights the need for more effective integration of external knowledge and reasoning mechanisms in LLMs to enhance their performance in real-world biomedical applications.

Neeladri Bhuiya, Viktor Schlegel, Stefan Winkler

State-of-the-art Large Language Models (LLMs) are accredited with an increasing number of different capabilities, ranging from reading comprehension, over advanced mathematical and reasoning skills to possessing scientific knowledge. In this paper we focus on their multi-hop reasoning capability: the ability to identify and integrate information from multiple textual sources. Given the concerns with the presence of simplifying cues in existing multi-hop reasoning benchmarks, which allow models to circumvent the reasoning requirement, we set out to investigate, whether LLMs are prone to exploiting such simplifying cues. We find evidence that they indeed circumvent the requirement to perform multi-hop reasoning, but they do so in more subtle ways than what was reported about their fine-tuned pre-trained language model (PLM) predecessors. Motivated by this finding, we propose a challenging multi-hop reasoning benchmark, by generating seemingly plausible multi-hop reasoning chains, which ultimately lead to incorrect answers. We evaluate multiple open and proprietary state-of-the-art LLMs, and find that their performance to perform multi-hop reasoning is affected, as indicated by up to 45% relative decrease in F1 score when presented with such seemingly plausible alternatives. We conduct a deeper analysis and find evidence that while LLMs tend to ignore misleading lexical cues, misleading reasoning paths indeed present a significant challenge.

Kuluhan Binici, Abhinav Ramesh Kashyap, Viktor Schlegel et al.

Automatic Speech Recognition (ASR) systems are pivotal in transcribing speech into text, yet the errors they introduce can significantly degrade the performance of downstream tasks like summarization. This issue is particularly pronounced in clinical dialogue summarization, a low-resource domain where supervised data for fine-tuning is scarce, necessitating the use of ASR models as black-box solutions. Employing conventional data augmentation for enhancing the noise robustness of summarization models is not feasible either due to the unavailability of sufficient medical dialogue audio recordings and corresponding ASR transcripts. To address this challenge, we propose MEDSAGE, an approach for generating synthetic samples for data augmentation using Large Language Models (LLMs). Specifically, we leverage the in-context learning capabilities of LLMs and instruct them to generate ASR-like errors based on a few available medical dialogue examples with audio recordings. Experimental results show that LLMs can effectively model ASR noise, and incorporating this noisy data into the training process significantly improves the robustness and accuracy of medical dialogue summarization systems. This approach addresses the challenges of noisy ASR outputs in critical applications, offering a robust solution to enhance the reliability of clinical dialogue summarization.

Viktor Schlegel, Goran Nenadic, Riza Batista-Navarro

Performance of NLP systems is typically evaluated by collecting a large-scale dataset by means of crowd-sourcing to train a data-driven model and evaluate it on a held-out portion of the data. This approach has been shown to suffer from spurious correlations and the lack of challenging examples that represent the diversity of natural language. Instead, we examine a framework for evaluating optimised models in training-set free setting on synthetically generated challenge sets. We find that despite the simplicity of the generation method, the data can compete with crowd-sourced datasets with regard to naturalness and lexical diversity for the purpose of evaluating the linguistic capabilities of MRC models. We conduct further experiments and show that state-of-the-art language model-based MRC systems can learn to succeed on the challenge set correctly, although, without capturing the general notion of the evaluated phenomenon.

Hao Li, Yuping Wu, Viktor Schlegel et al.

With the recent advances of large language models (LLMs), it is no longer infeasible to build an automated debate system that helps people to synthesise persuasive arguments. Previous work attempted this task by integrating multiple components. In our work, we introduce an argument mining dataset that captures the end-to-end process of preparing an argumentative essay for a debate, which covers the tasks of claim and evidence identification (Task 1 ED), evidence convincingness ranking (Task 2 ECR), argumentative essay summarisation and human preference ranking (Task 3 ASR) and metric learning for automated evaluation of resulting essays, based on human feedback along argument quality dimensions (Task 4 SQE). Our dataset contains 14k examples of claims that are fully annotated with the various properties supporting the aforementioned tasks. We evaluate multiple generative baselines for each of these tasks, including representative LLMs. We find, that while they show promising results on individual tasks in our benchmark, their end-to-end performance on all four tasks in succession deteriorates significantly, both in automated measures as well as in human-centred evaluation. This challenge presented by our proposed dataset motivates future research on end-to-end argument mining and summarisation. The repository of this project is available at https://github.com/HaoBytes/ArgSum-Datatset

Hao Li, Yu-Hao Huang, Chang Xu et al.

Time-series Generation (TSG) is a prominent research area with broad applications in simulations, data augmentation, and counterfactual analysis. While existing methods have shown promise in unconditional single-domain TSG, real-world applications demand for cross-domain approaches capable of controlled generation tailored to domain-specific constraints and instance-level requirements. In this paper, we argue that text can provide semantic insights, domain information and instance-specific temporal patterns, to guide and improve TSG. We introduce ``Text-Controlled TSG'', a task focused on generating realistic time series by incorporating textual descriptions. To address data scarcity in this setting, we propose a novel LLM-based Multi-Agent framework that synthesizes diverse, realistic text-to-TS datasets. Furthermore, we introduce BRIDGE, a hybrid text-controlled TSG framework that integrates semantic prototypes with text description for supporting domain-level guidance. This approach achieves state-of-the-art generation fidelity on 11 of 12 datasets, and improves controllability by up to 12% on MSE and 6% MAE compared to no text input generation, highlighting its potential for generating tailored time-series data.

Hao Li, Viktor Schlegel, Yizheng Sun et al.

Argument Mining (AM), a critical subfield of Natural Language Processing (NLP), focuses on extracting argumentative structures from text. The advent of Large Language Models (LLMs) has profoundly transformed AM, enabling advanced in-context learning, prompt-based generation, and robust cross-domain adaptability. This survey systematically synthesizes recent advancements in LLM-driven AM. We provide a concise review of foundational theories and annotation frameworks, alongside a meticulously curated catalog of datasets. A key contribution is our comprehensive taxonomy of AM subtasks, elucidating how contemporary LLM techniques -- such as prompting, chain-of-thought reasoning, and retrieval augmentation -- have reconfigured their execution. We further detail current LLM architectures and methodologies, critically assess evaluation practices, and delineate pivotal challenges including long-context reasoning, interpretability, and annotation bottlenecks. Conclusively, we highlight emerging trends and propose a forward-looking research agenda for LLM-based computational argumentation, aiming to strategically guide researchers in this rapidly evolving domain.

Viktor Schlegel, Anil A Bharath, Zilong Zhao et al.

Privacy-preserving synthetic data offers a promising solution to harness segregated data in high-stakes domains where information is compartmentalized for regulatory, privacy, or institutional reasons. This survey provides a comprehensive framework for understanding the landscape of privacy-preserving synthetic data, presenting the theoretical foundations of generative models and differential privacy followed by a review of state-of-the-art methods across tabular data, images, and text. Our synthesis of evaluation approaches highlights the fundamental trade-off between utility for down-stream tasks and privacy guarantees, while identifying critical research gaps: the lack of realistic benchmarks representing specialized domains and insufficient empirical evaluations required to contextualise formal guarantees. Through empirical analysis of four leading methods on five real-world datasets from specialized domains, we demonstrate significant performance degradation under realistic privacy constraints ($ε\leq 4$), revealing a substantial gap between results reported on general domain benchmarks and performance on domain-specific data. %Our findings highlight key challenges including unaccounted privacy leakage, insufficient empirical verification of formal guarantees, and a critical deficit of realistic benchmarks. These challenges underscore the need for robust evaluation frameworks, standardized benchmarks for specialized domains, and improved techniques to address the unique requirements of privacy-sensitive fields such that this technology can deliver on its considerable potential.

Hao Li, Bowen Deng, Chang Xu et al.

A unified foundation model for medical time series -- pretrained on open access and ethics board-approved medical corpora -- offers the potential to reduce annotation burdens, minimize model customization, and enable robust transfer across clinical institutions, modalities, and tasks, particularly in data-scarce or privacy-constrained environments. However, existing generalist time series foundation models struggle to handle medical time series data due to their inherent challenges, including irregular intervals, heterogeneous sampling rates, and frequent missing values. To address these challenges, we introduce MIRA, a unified foundation model specifically designed for medical time series forecasting. MIRA incorporates a Continuous-Time Rotary Positional Encoding that enables fine-grained modeling of variable time intervals, a frequency-specific mixture-of-experts layer that routes computation across latent frequency regimes to further promote temporal specialization, and a Continuous Dynamics Extrapolation Block based on Neural ODE that models the continuous trajectory of latent states, enabling accurate forecasting at arbitrary target timestamps. Pretrained on a large-scale and diverse medical corpus comprising over 454 billion time points collect from publicly available datasets, MIRA achieves reductions in forecasting errors by an average of 10% and 7% in out-of-distribution and in-distribution scenarios, respectively, when compared to other zero-shot and fine-tuned baselines. We also introduce a comprehensive benchmark spanning multiple downstream clinical tasks, establishing a foundation for future research in medical time series modeling.

Paul Pöhl, Viktor Schlegel, Hao Li et al.

Generating synthetic ECG data has numerous applications in healthcare, from educational purposes to simulating scenarios and forecasting trends. While recent diffusion models excel at generating short ECG segments, they struggle with longer sequences needed for many clinical applications. This paper proposes a novel three-layer synthesis framework for generating realistic long-form ECG signals. We first generate high-fidelity single beats using a diffusion model, then synthesize inter-beat features preserving critical temporal dependencies, and finally assemble beats into coherent long sequences using feature-guided matching. Our comprehensive evaluation demonstrates that the resulting synthetic ECGs maintain both beat-level morphological fidelity and clinically relevant inter-beat relationships. In arrhythmia classification tasks, our long-form synthetic ECGs significantly outperform end-to-end long-form ECG generation using the diffusion model, highlighting their potential for increasing utility for downstream applications. The approach enables generation of unprecedented multi-minute ECG sequences while preserving essential diagnostic characteristics.

Darius Feher, Abdullah Khered, Hao Zhang et al.

In an era increasingly dominated by digital platforms, the spread of misinformation poses a significant challenge, highlighting the need for solutions capable of assessing information veracity. Our research contributes to the field of Explainable Artificial Antelligence (XAI) by developing transformer-based fact-checking models that contextualise and justify their decisions by generating human-accessible explanations. Importantly, we also develop models for automatic evaluation of explanations for fact-checking verdicts across different dimensions such as \texttt{(self)-contradiction}, \texttt{hallucination}, \texttt{convincingness} and \texttt{overall quality}. By introducing human-centred evaluation methods and developing specialised datasets, we emphasise the need for aligning Artificial Intelligence (AI)-generated explanations with human judgements. This approach not only advances theoretical knowledge in XAI but also holds practical implications by enhancing the transparency, reliability and users' trust in AI-driven fact-checking systems. Furthermore, the development of our metric learning models is a first step towards potentially increasing efficiency and reducing reliance on extensive manual assessment. Based on experimental results, our best performing generative model \textsc{ROUGE-1} score of 47.77, demonstrating superior performance in generating fact-checking explanations, particularly when provided with high-quality evidence. Additionally, the best performing metric learning model showed a moderately strong correlation with human judgements on objective dimensions such as \texttt{(self)-contradiction and \texttt{hallucination}, achieving a Matthews Correlation Coefficient (MCC) of around 0.7.}

Yuping Wu, Viktor Schlegel, Warren Del-Pinto et al.

Training data is fundamental to the success of modern machine learning models, yet in high-stakes domains such as healthcare, the use of real-world training data is severely constrained by concerns over privacy leakage. A promising solution to this challenge is the use of differentially private (DP) synthetic data, which offers formal privacy guarantees while maintaining data utility. However, striking the right balance between privacy protection and utility remains challenging in clinical note synthesis, given its domain specificity and the complexity of long-form text generation. In this paper, we present Term2Note, a methodology to synthesise long clinical notes under strong DP constraints. By structurally separating content and form, Term2Note generates section-wise note content conditioned on DP medical terms, with each governed by separate DP constraints. A DP quality maximiser further enhances synthetic notes by selecting high-quality outputs. Experimental results show that Term2Note produces synthetic notes with statistical properties closely aligned with real clinical notes, demonstrating strong fidelity. In addition, multi-label classification models trained on these synthetic notes perform comparably to those trained on real data, confirming their high utility. Compared to existing DP text generation baselines, Term2Note achieves substantial improvements in both fidelity and utility while operating under fewer assumptions, suggesting its potential as a viable privacy-preserving alternative to using sensitive clinical notes.

Yidan Sun, Viktor Schlegel, Srinivasan Nandakumar et al.

Generative AI offers transformative potential for high-stakes domains such as healthcare and finance, yet privacy and regulatory barriers hinder the use of real-world data. To address this, differentially private synthetic data generation has emerged as a promising alternative. In this work, we introduce a unified benchmark to systematically evaluate the utility and fidelity of text datasets generated under formal Differential Privacy (DP) guarantees. Our benchmark addresses key challenges in domain-specific benchmarking, including choice of representative data and realistic privacy budgets, accounting for pre-training and a variety of evaluation metrics. We assess state-of-the-art privacy-preserving generation methods across five domain-specific datasets, revealing significant utility and fidelity degradation compared to real data, especially under strict privacy constraints. These findings underscore the limitations of current approaches, outline the need for advanced privacy-preserving data sharing methods and set a precedent regarding their evaluation in realistic scenarios.

Hao Li, Yizheng Sun, Viktor Schlegel et al.

Argument summarization aims to generate concise, structured representations of complex, multi-perspective debates. While recent work has advanced the identification and clustering of argumentative components, the generation stage remains underexplored. Existing approaches typically rely on single-pass generation, offering limited support for factual correction or structural refinement. To address this gap, we introduce Arg-LLaDA, a novel large language diffusion framework that iteratively improves summaries via sufficiency-guided remasking and regeneration. Our method combines a flexible masking controller with a sufficiency-checking module to identify and revise unsupported, redundant, or incomplete spans, yielding more faithful, concise, and coherent outputs. Empirical results on two benchmark datasets demonstrate that Arg-LLaDA surpasses state-of-the-art baselines in 7 out of 10 automatic evaluation metrics. In addition, human evaluations reveal substantial improvements across core dimensions, coverage, faithfulness, and conciseness, validating the effectiveness of our iterative, sufficiency-aware generation strategy.

Yulong Wu, Viktor Schlegel, Riza Batista-Navarro

As neural language models achieve human-comparable performance on Machine Reading Comprehension (MRC) and see widespread adoption, ensuring their robustness in real-world scenarios has become increasingly important. Current robustness evaluation research, though, primarily develops synthetic perturbation methods, leaving unclear how well they reflect real life scenarios. Considering this, we present a framework to automatically examine MRC models on naturally occurring textual perturbations, by replacing paragraph in MRC benchmarks with their counterparts based on available Wikipedia edit history. Such perturbation type is natural as its design does not stem from an arteficial generative process, inherently distinct from the previously investigated synthetic approaches. In a large-scale study encompassing SQUAD datasets and various model architectures we observe that natural perturbations result in performance degradation in pre-trained encoder language models. More worryingly, these state-of-the-art Flan-T5 and Large Language Models (LLMs) inherit these errors. Further experiments demonstrate that our findings generalise to natural perturbations found in other more challenging MRC benchmarks. In an effort to mitigate these errors, we show that it is possible to improve the robustness to natural perturbations by training on naturally or synthetically perturbed examples, though a noticeable gap still remains compared to performance on unperturbed data.

Vijay Prakash Dwivedi, Viktor Schlegel, Andy T. Liu et al.

Large Language Models (LLMs) have demonstrated remarkable performance across various domains, including healthcare. However, their ability to effectively represent structured non-textual data, such as the alphanumeric medical codes used in records like ICD-10 or SNOMED-CT, is limited and has been particularly exposed in recent research. This paper examines the challenges LLMs face in processing medical codes due to the shortcomings of current tokenization methods. As a result, we introduce the UniStruct architecture to design a multimodal medical foundation model of unstructured text and structured data, which addresses these challenges by adapting subword tokenization techniques specifically for the structured medical codes. Our approach is validated through model pre-training on both an extensive internal medical database and a public repository of structured medical records. Trained on over 1 billion tokens on the internal medical database, the proposed model achieves up to a 23% improvement in evaluation metrics, with around 2% gain attributed to our proposed tokenization. Additionally, when evaluated on the EHRSHOT public benchmark with a 1/1000 fraction of the pre-training data, the UniStruct model improves performance on over 42% of the downstream tasks. Our approach not only enhances the representation and generalization capabilities of patient-centric models but also bridges a critical gap in representation learning models' ability to handle complex structured medical data, alongside unstructured text.

Viktor Schlegel, Abhinav Ramesh Kashyap, Thanh-Tung Nguyen et al.

Computerised clinical coding approaches aim to automate the process of assigning a set of codes to medical records. While there is active research pushing the state of the art on clinical coding for hospitalized patients, the outpatient setting -- where doctors tend to non-hospitalised patients -- is overlooked. Although both settings can be formalised as a multi-label classification task, they present unique and distinct challenges, which raises the question of whether the success of inpatient clinical coding approaches translates to the outpatient setting. This paper is the first to investigate how well state-of-the-art deep learning-based clinical coding approaches work in the outpatient setting at hospital scale. To this end, we collect a large outpatient dataset comprising over 7 million notes documenting over half a million patients. We adapt four state-of-the-art clinical coding approaches to this setting and evaluate their potential to assist coders. We find evidence that clinical coding in outpatient settings can benefit from more innovations in popular inpatient coding benchmarks. A deeper analysis of the factors contributing to the success -- amount and form of data and choice of document representation -- reveals the presence of easy-to-solve examples, the coding of which can be completely automated with a low error rate.

Yidan Sun, Viktor Schlegel, Srinivasan Nandakumar et al.

Data-driven decision support in high-stakes domains like healthcare and finance faces significant barriers to data sharing due to regulatory, institutional, and privacy concerns. While recent generative AI models, such as large language models, have shown impressive performance in open-domain tasks, their adoption in sensitive environments remains limited by unpredictable behaviors and insufficient privacy-preserving datasets for benchmarking. Existing anonymization methods are often inadequate, especially for unstructured text, as redaction and masking can still allow re-identification. Differential Privacy (DP) offers a principled alternative, enabling the generation of synthetic data with formal privacy assurances. In this work, we address these challenges through three key contributions. First, we introduce a comprehensive evaluation framework with standardized utility and fidelity metrics, encompassing nine curated datasets that capture domain-specific complexities such as technical jargon, long-context dependencies, and specialized document structures. Second, we conduct a large-scale empirical study benchmarking state-of-the-art DP text generation methods and LLMs of varying sizes and different fine-tuning strategies, revealing that high-quality domain-specific synthetic data generation under DP constraints remains an unsolved challenge, with performance degrading as domain complexity increases. Third, we develop a membership inference attack (MIA) methodology tailored for synthetic text, providing first empirical evidence that the use of public datasets - potentially present in pre-training corpora - can invalidate claimed privacy guarantees. Our findings underscore the urgent need for rigorous privacy auditing and highlight persistent gaps between open-domain and specialist evaluations, informing responsible deployment of generative AI in privacy-sensitive, high-stakes settings.

Mingyang Li, Viktor Schlegel, Tingting Mu et al.

Mapping clinical documents to standardised clinical vocabularies is an important task, as it provides structured data for information retrieval and analysis, which is essential to clinical research, hospital administration and improving patient care. However, manual coding is both difficult and time-consuming, making it impractical at scale. Automated coding can potentially alleviate this burden, improving the availability and accuracy of structured clinical data. The task is difficult to automate, as it requires mapping to high-dimensional and long-tailed target spaces, such as the International Classification of Diseases (ICD). While external knowledge sources have been readily utilised to enhance output code representation, the use of external resources for representing the input documents has been underexplored. In this work, we compute a structured representation of the input documents, making use of document-level knowledge graphs (KGs) that provide a comprehensive structured view of a patient's condition. The resulting knowledge graph efficiently represents the patient-centred input documents with 23\% of the original text while retaining 90\% of the information. We assess the effectiveness of this graph for automated ICD-9 coding by integrating it into the state-of-the-art ICD coding architecture PLM-ICD. Our experiments yield improved Macro-F1 scores by up to 3.20\% on popular benchmarks, while improving training efficiency. We attribute this improvement to different types of entities and relationships in the KG, and demonstrate the improved explainability potential of the approach over the text-only baseline.

Yulong Wu, Viktor Schlegel, Riza Batista-Navarro

How does the natural evolution of context paragraphs affect question answering in generative Large Language Models (LLMs)? To investigate this, we propose a framework for curating naturally evolved, human-edited variants of reading passages from contemporary QA benchmarks and for analyzing LLM performance across a range of semantic similarity scores, which quantify how closely each variant aligns with content seen during pretraining. Using this framework, we evaluate six QA datasets and eight LLMs with publicly available training data. Our experiments reveal that LLM performance declines as reading passages naturally diverge from the versions encountered during pretraining-even when the question and all necessary information remains present at inference time. For instance, average model accuracy on BoolQ drops by over 30% from the highest to lowest similarity bins, with slopes exceeding 70 across several LLMs. These findings suggest that natural text evolution poses a significant challenge to the language understanding capabilities of LLMs.

Mingyang Li, Viktor Schlegel, Tingting Mu et al.

Automated clinical coding involves mapping unstructured text from Electronic Health Records (EHRs) to standardized code systems such as the International Classification of Diseases (ICD). While recent advances in deep learning have significantly improved the accuracy and efficiency of ICD coding, the lack of explainability in these models remains a major limitation, undermining trust and transparency. Current explorations about explainability largely rely on attention-based techniques and qualitative assessments by physicians, yet lack systematic evaluation using consistent criteria on high-quality rationale datasets, as well as dedicated approaches explicitly trained to generate rationales for further enhancing explanation. In this work, we conduct a comprehensive evaluation of the explainability of the rationales for ICD coding through two key lenses: faithfulness that evaluates how well explanations reflect the model's actual reasoning and plausibility that measures how consistent the explanations are with human expert judgment. To facilitate the evaluation of plausibility, we construct a new rationale-annotated dataset, offering denser annotations with diverse granularity and aligns better with current clinical practice, and conduct evaluation across three types of rationales of ICD coding. Encouraged by the promising plausibility of LLM-generated rationales for ICD coding, we further propose new rationale learning methods to improve the quality of model-generated rationales, where rationales produced by prompting LLMs with/without annotation examples are used as distant supervision signals. We empirically find that LLM-generated rationales align most closely with those of human experts. Moreover, incorporating few-shot human-annotated examples not only further improves rationale generation but also enhances rationale-learning approaches.

Anand Subramanian, Viktor Schlegel, Abhinav Ramesh Kashyap et al.

There is vivid research on adapting Large Language Models (LLMs) to perform a variety of tasks in high-stakes domains such as healthcare. Despite their popularity, there is a lack of understanding of the extent and contributing factors that allow LLMs to recall relevant knowledge and combine it with presented information in the clinical and biomedical domain: a fundamental pre-requisite for success on down-stream tasks. Addressing this gap, we use Multiple Choice and Abstractive Question Answering to conduct a large-scale empirical study on 22 datasets in three generalist and three specialist biomedical sub-domains. Our multifaceted analysis of the performance of 15 LLMs, further broken down by sub-domain, source of knowledge and model architecture, uncovers success factors such as instruction tuning that lead to improved recall and comprehension. We further show that while recently proposed domain-adapted models may lack adequate knowledge, directly fine-tuning on our collected medical knowledge datasets shows encouraging results, even generalising to unseen specialist sub-domains. We complement the quantitative results with a skill-oriented manual error analysis, which reveals a significant gap between the models' capabilities to simply recall necessary knowledge and to integrate it with the presented context. To foster research and collaboration in this field we share M-QALM, our resources, standardised methodology, and evaluation results, with the research community to facilitate further advancements in clinical knowledge representation learning within language models.

Thanh-Tung Nguyen, Viktor Schlegel, Abhinav Kashyap et al.

Clinical notes in healthcare facilities are tagged with the International Classification of Diseases (ICD) code; a list of classification codes for medical diagnoses and procedures. ICD coding is a challenging multilabel text classification problem due to noisy clinical document inputs and long-tailed label distribution. Recent automated ICD coding efforts improve performance by encoding medical notes and codes with additional data and knowledge bases. However, most of them do not reflect how human coders generate the code: first, the coders select general code categories and then look for specific subcategories that are relevant to a patient's condition. Inspired by this, we propose a two-stage decoding mechanism to predict ICD codes. Our model uses the hierarchical properties of the codes to split the prediction into two steps: At first, we predict the parent code and then predict the child code based on the previous prediction. Experiments on the public MIMIC-III data set show that our model performs well in single-model settings without external data or knowledge.

Hao Li, Viktor Schlegel, Riza Batista-Navarro et al.

Argument summarisation is a promising but currently under-explored field. Recent work has aimed to provide textual summaries in the form of concise and salient short texts, i.e., key points (KPs), in a task known as Key Point Analysis (KPA). One of the main challenges in KPA is finding high-quality key point candidates from dozens of arguments even in a small corpus. Furthermore, evaluating key points is crucial in ensuring that the automatically generated summaries are useful. Although automatic methods for evaluating summarisation have considerably advanced over the years, they mainly focus on sentence-level comparison, making it difficult to measure the quality of a summary (a set of KPs) as a whole. Aggravating this problem is the fact that human evaluation is costly and unreproducible. To address the above issues, we propose a two-step abstractive summarisation framework based on neural topic modelling with an iterative clustering procedure, to generate key points which are aligned with how humans identify key points. Our experiments show that our framework advances the state of the art in KPA, with performance improvement of up to 14 (absolute) percentage points, in terms of both ROUGE and our own proposed evaluation metrics. Furthermore, we evaluate the generated summaries using a novel set-based evaluation toolkit. Our quantitative analysis demonstrates the effectiveness of our proposed evaluation metrics in assessing the quality of generated KPs. Human evaluation further demonstrates the advantages of our approach and validates that our proposed evaluation metric is more consistent with human judgment than ROUGE scores.

Abhinav Ramesh Kashyap, Thanh-Tung Nguyen, Viktor Schlegel et al.

Sentence representations are a critical component in NLP applications such as retrieval, question answering, and text classification. They capture the meaning of a sentence, enabling machines to understand and reason over human language. In recent years, significant progress has been made in developing methods for learning sentence representations, including unsupervised, supervised, and transfer learning approaches. However there is no literature review on sentence representations till now. In this paper, we provide an overview of the different methods for sentence representation learning, focusing mostly on deep learning models. We provide a systematic organization of the literature, highlighting the key contributions and challenges in this area. Overall, our review highlights the importance of this area in natural language processing, the progress made in sentence representation learning, and the challenges that remain. We conclude with directions for future research, suggesting potential avenues for improving the quality and efficiency of sentence representations.

Viktor Schlegel, Goran Nenadic, Riza Batista-Navarro

Advances in NLP have yielded impressive results for the task of machine reading comprehension (MRC), with approaches having been reported to achieve performance comparable to that of humans. In this paper, we investigate whether state-of-the-art MRC models are able to correctly process Semantics Altering Modifications (SAM): linguistically-motivated phenomena that alter the semantics of a sentence while preserving most of its lexical surface form. We present a method to automatically generate and align challenge sets featuring original and altered examples. We further propose a novel evaluation methodology to correctly assess the capability of MRC systems to process these examples independent of the data they were optimised on, by discounting for effects introduced by domain shift. In a large-scale empirical study, we apply the methodology in order to evaluate extractive MRC models with regard to their capability to correctly process SAM-enriched data. We comprehensively cover 12 different state-of-the-art neural architecture configurations and four training datasets and find that -- despite their well-known remarkable performance -- optimised models consistently struggle to correctly process semantically altered data.

Viktor Schlegel, Goran Nenadic, Riza Batista-Navarro

Recent years have seen a growing number of publications that analyse Natural Language Inference (NLI) datasets for superficial cues, whether they undermine the complexity of the tasks underlying those datasets and how they impact those models that are optimised and evaluated on this data. This structured survey provides an overview of the evolving research area by categorising reported weaknesses in models and datasets and the methods proposed to reveal and alleviate those weaknesses for the English language. We summarise and discuss the findings and conclude with a set of recommendations for possible future research directions. We hope it will be a useful resource for researchers who propose new datasets, to have a set of tools to assess the suitability and quality of their data to evaluate various phenomena of interest, as well as those who develop novel architectures, to further understand the implications of their improvements with respect to their model's acquired capabilities.

Viktor Schlegel, Marco Valentino, André Freitas et al.

Machine Reading Comprehension (MRC) is the task of answering a question over a paragraph of text. While neural MRC systems gain popularity and achieve noticeable performance, issues are being raised with the methodology used to establish their performance, particularly concerning the data design of gold standards that are used to evaluate them. There is but a limited understanding of the challenges present in this data, which makes it hard to draw comparisons and formulate reliable hypotheses. As a first step towards alleviating the problem, this paper proposes a unifying framework to systematically investigate the present linguistic features, required reasoning and background knowledge and factual correctness on one hand, and the presence of lexical cues as a lower bound for the requirement of understanding on the other hand. We propose a qualitative annotation schema for the first and a set of approximative metrics for the latter. In a first application of the framework, we analyse modern MRC gold standards and present our findings: the absence of features that contribute towards lexical ambiguity, the varying factual correctness of the expected answers and the presence of lexical cues, all of which potentially lower the reading comprehension complexity and quality of the evaluation data.

Mokanarangan Thayaparan, Marco Valentino, Viktor Schlegel et al.

Recent advances in reading comprehension have resulted in models that surpass human performance when the answer is contained in a single, continuous passage of text. However, complex Question Answering (QA) typically requires multi-hop reasoning - i.e. the integration of supporting facts from different sources, to infer the correct answer. This paper proposes Document Graph Network (DGN), a message passing architecture for the identification of supporting facts over a graph-structured representation of text. The evaluation on HotpotQA shows that DGN obtains competitive results when compared to a reading comprehension baseline operating on raw text, confirming the relevance of structured representations for supporting multi-hop reasoning.