Multiple Versions Research Articles

Faktor - Faktor yang Mempengaruhi Kepatuhan Wajib Pajak UMKM dalam Melaksanakan Kewajiban Perpajakan: Studi Kasus pada Wajib Pajak UMKM di KPP Pratama Seberang Ulu

This study aims to examine the effect of tax knowledge, tax sanctions, service quality, tax rates and application of e-filing on taxpayer compliance of UMKMs at KPP Pratama Seberang Ulu. The independent variables in this study are tax knowledge, tax sanctions, service quality, tax rates and application of e-filing while the dependent variable is taxpayer compliance. There are 2 theories in this study, namely attribution and the theory of Planned Behavior. This study uses a quantitative method. The population in this study are UMKM actors registered at KPP Pratama Seberang Ulu. The sample was determined using purposive sampling. Data were collected by distributing questionnaires to 47 UMKMs registered at KPP Pratama Seberang Ulu and in accordance with the specified characteristics. Primary data collection techniques and data analysis using multiple linear regression analysis SPSS software version 25. The methods used in this study are Descriptive Analysis, Validity Test, Reliability Test, Classical Assumption Test, Multiple Linear Regression Analysis Test, F Test, R2 Determination Coefficient Test and t Test. The results of the study indicate that tax knowledge, tax sanctions and the application of e-filing have an effect on taxpayer compliance, while the quality of service and tax rates have no effect on the compliance of UMKM taxpayers at the Seberang Ulu Pratama Tax Office..

Lightweight convolutional neural network for chest X-ray images classification

In this study, we developed a lightweight and rapid convolutional neural network (CNN) architecture for chest X-ray images; it primarily consists of a redesigned feature extraction (FE) module and multiscale feature (MF) module and validated using publicly available COVID-19 datasets. Experiments were conducted on multiple updated versions of the COVID-19 Radiography Database, a publicly accessible dataset on Kaggle. The database contained images categorized into three classes: COVID-19 coronavirus, viral or bacterial pneumonia, and normal. The results revealed that the proposed method achieved a training accuracy of 99.85% and a validation accuracy of 96.28% when detecting the three classes. In the test set, the optimal results were 96.03% accuracy for COVID-19, 97.10% accuracy for viral or bacterial pneumonia, and 97.86% accuracy for normal individuals. By reducing the computational requirements and improving the speed of the model, the proposed method can achieve real-time, low-error performance to help medical professionals with rapid diagnosis of COVID-19.

HawkEye Conv-Driven YOLOv10 with Advanced Feature Pyramid Networks for Small Object Detection in UAV Imagery

Current mainstream computer vision algorithms focus on designing suitable network architectures and loss functions to fit training data. However, the accuracy of small object detection remains lower than for other scales, and the design of convolution operators limits the model’s performance. For UAV small object detection, standard convolutions, due to their fixed kernel size, cannot adaptively capture small object spatial information. Many convolutional variants have scattered sampling points, leading to blurred boundaries and reduced accuracy. In response, we propose HawkEye Conv (HEConv), which utilizes stable sampling and dynamic offsets with random selection. By varying the convolution kernel design, HEConv reduces the accuracy gap between small and larger objects while offering multiple versions and plug-and-play capabilities. We also develop HawkEye Spatial Pyramid Pooling and Gradual Dynamic Feature Pyramid Network modules to validate HEConv. Experiments on the RFRB agricultural and VisDrone2019 urban datasets demonstrate that, compared to YOLOv10, our model improves AP50 by 11.9% and 6.2%, APS by 11.5% and 5%, and F1-score by 5% and 7%. Importantly, it enhances small object detection without sacrificing large object accuracy, thereby reducing the multi-scale performance gap.

Structural characterization and AlphaFold modeling of human T cell receptor recognition of NRAS cancer neoantigens.

T cell receptors (TCRs) that recognize cancer neoantigens are important for anticancer immune responses and immunotherapy. Understanding the structural basis of TCR recognition of neoantigens provides insights into their exquisite specificity and can enable design of optimized TCRs. We determined crystal structures of a human TCR in complex with NRAS Q61K and Q61R neoantigen peptides and HLA-A1 major histocompatibility complex (MHC), revealing the molecular underpinnings for dual recognition and specificity versus wild-type NRAS peptide. We then used multiple versions of AlphaFold to model the corresponding complex structures, given the challenge of immune recognition for such methods. One implementation of AlphaFold2 (TCRmodel2) with additional sampling was able to generate accurate models of the complexes, while AlphaFold3 also showed strong performance, although success was lower for other complexes. This study provides insights into TCR recognition of a shared cancer neoantigen as well as the utility and practical considerations for using AlphaFold to model TCR-peptide-MHC complexes.

Technical variability across the 450K, EPICv1, and EPICv2 DNA methylation arrays: lessons learned for clinical and longitudinal studies

DNA methylation (DNAm) is the most commonly measured epigenetic mechanism in human populations, with most studies using Illumina arrays to assess DNAm levels. In 2023, Illumina updated their DNAm arrays to the EPIC version 2 (EPICv2), building on prior iterations, namely the EPIC version 1 (EPICv1) and 450K arrays. Whether DNAm measurements are stable across these three generations of arrays has yet not been investigated, limiting the ability of researchers—especially those with longitudinal data—to compare and replicate results across arrays. Here, we present results from a study of 30 child participants (15 male; 15 female) from the Drakenstein Child Health Study, who had DNAm measured on all three of the latest arrays: 450K, EPICv1, and EPICv2. Using these data, we created an annotation of probe quality across arrays, which includes the intraclass correlations, interquartile ranges, correlations, and array bias (i.e., the extent to which DNAm levels were explained by array type) of all CpGs. We also present results from an analysis of sex differences, where we found that CpGs with lower replicability across arrays had higher array-based variance, suggesting this variance metric help guide replication efforts. We also showed that epigenetic age estimates across arrays were more stable when using the principal component versions of epigenetic clocks. Ultimately, this collection of results provides a framework for investigating the replicability and longitudinal stability of epigenetic changes across multiple versions of Illumina DNAm arrays.

Continual learning in the presence of repetition

Continual learning (CL) provides a framework for training models in ever-evolving environments. Although re-occurrence of previously seen objects or tasks is common in real-world problems, the concept of repetition in the data stream is not often considered in standard benchmarks for CL. Unlike with the rehearsal mechanism in buffer-based strategies, where sample repetition is controlled by the strategy, repetition in the data stream naturally stems from the environment. This report provides a summary of the CLVision challenge at CVPR 2023, which focused on the topic of repetition in class-incremental learning. The report initially outlines the challenge objective and then describes three solutions proposed by finalist teams that aim to effectively exploit the repetition in the stream to learn continually. The experimental results from the challenge highlight the effectiveness of ensemble-based solutions that employ multiple versions of similar modules, each trained on different but overlapping subsets of classes. This report underscores the transformative potential of taking a different perspective in CL by employing repetition in the data stream to foster innovative strategy design.

Incremental model transformations with triple graph grammars for multi-version models and multi-version pattern matching

AbstractLike conventional software projects, projects in model-driven software engineering require adequate management of multiple versions of development artifacts, importantly allowing living with temporary inconsistencies. In previous work, we have introduced multi-version models for model-driven software engineering, which allow checking well-formedness and finding merge conflicts for multiple versions of the same model at once. However, situations where different models are linked via automatic model transformations also have to be handled for multi-version models. In this paper, we propose a technique for jointly handling the transformation of multiple versions of a source model into corresponding versions of a target model. This enables the use of a more compact representation that may afford improved execution time of both the transformation and further analysis. Our approach is based on the well-known formalism of triple graph grammars and the aforementioned encoding of model version histories called multi-version models. In addition to batch transformation of an entire history, the technique covers incremental synchronization of changes in the framework of multi-version models. Our solution is complemented by a dedicated pattern matching technique for multi-version models. We show the correctness of our approach with respect to the standard semantics of triple graph grammars and conduct an empirical evaluation to investigate the performance of our technique regarding execution time and memory consumption. Our results indicate that the proposed solution affords lower memory consumption and may improve execution time for batch transformation of large version histories, but can also come with computational overhead in unfavorable cases.

Analysis of Land Surface Performance Differences and Uncertainty in Multiple Versions of MODIS LST Products

Moderate Resolution Imaging Spectroradiometer (MODIS) Land Surface Temperature (LST) products are essential data sources for global and regional climate change research. Currently, several versions of the MODIS LST product have been released, yet the performance differences and uncertainties they introduce in land surface studies remain insufficiently addressed. To bridge this gap, this study focuses on four distinct versions of the LST product: MxD11A1 Collection 5 (C5), Collection 6 (C6), Collection 6.1 (C6.1), and MxD21A1 Collection 6.1 (MxD21). The spatial resolution of all product generations is 1 km, and the temporal resolution is 0.5 days. This study provides a comprehensive analysis of the errors arising from different generations of these products in various land surface process studies. The error assessment includes cross-comparisons between product versions and evaluations of the absolute errors generated. Absolute errors in evaluation data were collected from 13 surface sites within the Heihe Watershed Allied Telemetry Experimental Research (HiWATER) project during the period 2013–2018. Cross-validation results show that the largest difference between C5 and C6.1 occurs over bare land, with an RMSE of approximately 1.45 K, while there is no significant change between C6 and C6.1. MOD21 shows considerable variation compared to C6.1 at night across different land cover types, with RMSE over cropland exceeding 2 K. The temperature difference between MOD21 and C6.1 is more pronounced at night (2.01 K) than during the day (0.30 K). Validation results based on temperature indicate that C5 has greater uncertainty compared to C6, especially over bare land, where errors are 2.06 K and 1.06 K, respectively. Furthermore, MxD21 demonstrates significant day–night performance discrepancies, with an average bias of 0.10 K at night, while daytime errors over bare land can reach 2 K, potentially influenced by atmospheric conditions. Based on the research in this paper, it is possible to clarify the performance of different versions of MODIS products, reflecting the appropriateness of their past applications; on the other hand, it is recommended to prioritize the use of the MxD11A1 C6 and C6.1 products for monitoring and applications in bare soil areas to ensure higher accuracy. Furthermore, for day and night monitoring, it may be beneficial to alternate between the MxD11A1 and MxD21A1 products to fully leverage their respective advantages and enhance overall monitoring effectiveness.

Scalable multimodal approach for face generation and super-resolution using a conditional diffusion model

Multimodal Conditioned face image generation and face super-resolution are significant areas of research. To achieve optimal results, this paper utilizes diffusion models as the primary engine for these tasks. This paper presents two main contributions: (1) “Speaking the Language of Faces” (SLF): a flexible, modular, fusion-less and architecturally simple multimodal system. (2) A Scalability scheme and a sensitivity analysis which can assist practitioners in system parameter estimation and feature selection. SLF consists of two main components: a feature vector generator (encoder), and an image generator (decoder) utilizing a conditional diffusion model. SLF can accept various inputs, including low-resolution images, speech signals, person attributes (age, gender, ethnicity), or any combination of these. Moreover, Scalability based on conditional scale values is utilized. The implementation of SLF has confirmed its versatility (e.g., speech to face image generation, conditioned face super-resolution). We trained multiple system versions to conduct a sensitivity analysis and to determine the influence of each individual feature on the output image. Consequently, speaker embeddings have proven to be sufficient audio features for our task. It was also found that the effects of audio signals are profound and are more pronounced than those of the low resolution images (8 × 8), whose effects are still significant. The effect of gender, ethnicity and age were found to be moderate. On another note, conditional scale values significantly impact the system’s behavior and performance.

A Study on Hadith Authenticity Concerning the Abrogation of Bequests for Iddah Maintenance by Heirs and Its Relevance to Islamic Family Law in Indonesia

Although Islamic jurists (fiqh scholars) have reached a consensus on the elimination of the provision for iddah maintenance (post-divorce or mourning period support) through inheritance shares upon a husband�s death, the hadith cited as the basis for this elimination remains a subject of debate. This consensus carries sociological, legal, and theological implications that may disadvantage wives, particularly in the aftermath of their husbands' deaths. The purpose of this study is to reassess the authenticity of the hadith concerning the elimination of iddah maintenance via inheritance distribution. This paper employs Abou El Fadl's hermeneutic negotiation approach to scrutinize the authenticity of the hadith regarding the elimination of iddah maintenance and to contextualize it within the Indonesian framework. The findings indicate that the hadith serving as the foundation for this issue is an attributed to a companion rather than the Prophet (mauquf hadith), whose chain of narration (sanad) is contested. Certain narrators are considered unreliable, and there are discrepancies within its text (matn), with multiple versions existing. Consequently, this paper refrains from relying on the doctrine of abrogation (naskh) and supports the view that iddah maintenance following a husband�s death is an option that need not conflict with the wife�s inheritance share. Both iddah maintenance and the wife�s inheritance share can thus be implemented concurrently. The paper posits that, in the Indonesian context, iddah maintenance upon the husband�s death should be regarded as a form of joint property that can be allocated without reducing the wife�s inheritance share. The implementation of both joint property distribution and inheritance safeguards the rights of the wife and her children following the husband's death.

Reflecting on the Impact of Generative AI for Sustainability Accounting Scholarship

ABSTRACT Our yearly editorial is dedicated to reflections concerning two recent digitalisation-based developments, both of which are currently radically transforming the publishing arena. We begin by considering the rapid emergence of generative Artificial Intelligence (AI) models known as Large Language Models (LLMs), which have already had tremendous effects on academic knowledge production. In this context, we clarify our stance on the use of AI in articles submitted to Social & Environmental Accountability Journal (SEAJ) and reviews conducted for the journal. We then continue by taking a closer look at the proliferation of systematic literature reviews, in their multiple versions, e.g. bibliometric-based reviews, which are submitted to the journal in increasing numbers. We use this editorial to revisit our editorial policy and outline the kind of submissions welcomed at SEAJ.

ENHANCING OTFS MODULATION FOR 6G THROUGH HYBRID PAPR REDUCTION TECHNIQUE FOR DIFFERENT SUB-CARRIERS

Peak-to-average power ratio (PAPR) in orthogonal time frequency space (OTFS) is vital for optimizing signal efficiency, minimizing distortion, and enhancing the performance of a beyond fifth-generation (B5G) system. The paper focuses on lowering the PAPR while retaining the bit error rate (BER) and power spectral density (PSD) for 64 and 256 sub-carriers. The PAPR is minimized by using a fractal optimization process known as selective mapping (SLM) and partial transmission sequence (PTS) at the transmitting block of the OTFS framework. The combination of SLM and PTS reduces peak power by selectively modifying the phase of transmitted symbols. SLM generates multiple versions of the signal, while PTS allocates power to these versions optimally, collectively mitigating amplitude peaks and minimizing PAPR. According to the simulation results, the suggested SLM+PTS works better than the traditional SLM and PTS, achieving low spectrum leakage and PAPR improvements of 5.5 dB and 4.9[Formula: see text]dB as well as SNR gains of 2.4[Formula: see text]dB and 2.1[Formula: see text]dB. In the future, SLM+PTS may work on lowering PAPR by making computers faster, researching adaptive algorithms, and combining machine learning methods for real-time optimization in various communication settings. Complex modeling, including fractal-based SLM and PTS approaches, provides a detailed understanding of waveform behavior and interference patterns, enabling more accurate and effective hybrid PAPR reduction techniques. This combination enhances the ability to manage to mitigate non-linear distortion across various sub-carriers, crucial for the performance and efficiency of 6G OTFS modulation.

An international environmental scan of the scope and characteristics of patient decision aids which are freely available online

ObjectiveTo analyse the scope and characteristics of freely available online patient decision aids. MethodsAn international environmental scan of online decision aids, with no language restriction, was conducted by searching aids in the recent Cochrane review and 35 online sources. Aid characteristics were extracted and analysed. ResultsOf 1555 decision aids identified, there were 1222 unique aids after removing duplicates. Of these, 785 met eligibility criteria. The aids were from 17 countries (30 % from the Netherlands, 20 % from the US) and 44 % in English. Nearly all had a downloadable PDF or were printable and 32 % were interactive web-based. Many aids (72 %) were about a treatment decision. Most common diseases addressed were cancer (23 %), especially breast cancer, musculoskeletal diseases (11 %), and genitourinary system diseases (10 %). ConclusionsThere are many freely available decision aids, with most addressing treatment decisions and particular health conditions. Many (21 %) of the aids identified were duplicates, some topics are covered by multiple versions of decision aids, and other topics have few decision aids. Practice implicationsWhile many decision aids exist online, initiatives are needed to minimise aid duplication, achieve better distribution of aids across health decisions/topics, encourage greater collaboration between developers, and adaption of existing aids.

Weak saliency ensemble network for person Re-identification using infrared light images

In recent years, person re-identification (re-id) has primarily been studied using visible light (VL) images. However, the challenges of employing VL images in nighttime environments have prompted research into using infrared light (IR) images. Yet, the utilization of both VL and IR images in person re-id has resulted in increased computational cost and processing time in multi-modality systems, leading to studies focusing solely on IR images. Nevertheless, IR images, lacking color and texture information, generally yield lower recognition performance in existing person re-id studies. In addition, previous studies have shown that person re-id performance suffers in the presence of complex background noise. To tackle these challenges, this study proposes a new weak saliency ensemble network (WSE-Net) for person re-id using IR images. WSE-Net incorporates a channel reduction of feature (CRF) method to reduce computational cost in the ensemble network, a technique for converting input images into group of patch images and feeding them into the ensemble model to enhance the reduced feature information, and a grouped convolution ensemble network (GCE-Net) that enables the fusion of features extracted from original and attention-guided ensemble models.The performance of person re-id using WSE-Net was evaluated on the Dongguk body-based person recognition database version 1 (DBPerson-Recog-DB1) and the Sun Yat-sen university multiple modality re-identification version 1 (SYSU-MM01). Experimental results demonstrated that on DBPerson-Recog-DB1, WSE-Net achieved 93.65% in rank 1, 95.28% in mean average precision (mAP), and 93.52% in the harmonic mean of precision and recall. Additionally, on SYSU-MM01, WSE-Net achieved 86.85% in rank 1, 44.58% in mAP, and 40.06% in the harmonic mean of precision and recall. Furthermore, the accuracy of WSE-Net on both datasets surpassed that of state-of-the-art methods.

Sexual Harassment, Sexual Violence, and Mental Health Outcomes: Causal Inference with Ambiguous Exposures.

Social exposures and their impact on mental health has proven hard to capture, partly owing to the complex and multifaceted nature of social reality. Sexual harassment and sexual violence (SHV) are no exceptions. SHV can be conceptualized as a continuum of negative sexual experiences whose severity vary depending on multiple determinants. Further, SHV can be conceptualized as either discrete events or as a generally hostile sexual environment represented by latent variables. With any of these conceptualizations, SHV constitutes a broad construct containing many kinds of negative experiences. This ambiguity poses challenges for determining the mental health consequences, as different forms of SHV may vary in terms of their mental health impact. We discuss different conceptualizations of SHV in relation to mental health outcomes through the lens of the potential outcomes framework, with a focus on the consistency condition. The multiple versions of treatment theory is presented to show how to provide formal interpretations of causal estimates under ambiguous exposures. Lastly, we provide suggestions on how the increase the clarity and interpretability of the effects of SHV on mental health, by increasing the precision of the causal questions and the use of more specific definitions of SHV.

Mount SMB.pcap: Reconstructing file systems and file operations from network traffic

File system and network forensics are fundamental in forensic investigations, but are often treated as distinct disciplines. This work seeks to unify these fields by introducing a novel framework capable of mounting network captures, enabling investigators to seamlessly browse data using conventional tools. Although our implementation supports various protocols such as HTTP, TLS, and FTP, this work will particularly focus on the complexities of the Server Message Block (SMB) protocol, which is fundamental for shared file system access, especially within local networks.For this, we present a detailed methodology to extract essential file system data from SMB network traffic, aiming to reconstruct the share's file system as accurately as the original. Our approach goes beyond traditional tools like Wireshark, which typically only extract individual files from SMB transmissions. Instead, we reconstruct the entire file system hierarchy, retrieve all associated metadata, and handle multiple versions of files captured within the same network traffic. In addition, we also investigate how file operations impact SMB commands and show how these can be used to accurately recreate user activities on an SMB share based solely on network traffic. Although both methodologies and implementations can be applied independently, their combination provides investigators with a comprehensive view of the reconstructed file system along with the corresponding user activities extracted from network traffic.

Estimation of Fractal Dimension and Segmentation of Body Regions for Deep Learning-Based Gender Recognition

There are few studies utilizing only IR cameras for long-distance gender recognition, and they have shown low recognition performance due to their lack of color and texture information in IR images with a complex background. Therefore, a rough body segmentation-based gender recognition network (RBSG-Net) is proposed, with enhanced gender recognition performance achieved by emphasizing the silhouette of a person through a body segmentation network. Anthropometric loss for the segmentation network and an adaptive body attention module are also proposed, which effectively integrate the segmentation and classification networks. To enhance the analytic capabilities of the proposed framework, fractal dimension estimation was introduced into the system to gain insights into the complexity and irregularity of the body region, thereby predicting the accuracy of body segmentation. For experiments, near-infrared images from the Sun Yat-sen University multiple modality re-identification version 1 (SYSU-MM01) dataset and thermal images from the Dongguk body-based gender version 2 (DBGender-DB2) database were used. The equal error rates of gender recognition by the proposed model were 4.320% and 8.303% for these two databases, respectively, surpassing state-of-the-art methods.

On Newton–Cotes Formula-Type Inequalities for Multiplicative Generalized Convex Functions via Riemann–Liouville Fractional Integrals with Applications to Quadrature Formulas and Computational Analysis

In this article, we develop multiplicative fractional versions of Simpson’s and Newton’s formula-type inequalities for differentiable generalized convex functions with the help of established identities. The main motivation for using generalized convex functions lies in their ability to extend results beyond traditional convex functions, encompassing a broader class of functions, and providing optimal approximations for both lower and upper bounds. These inequalities are very useful in finding the error bounds for the numerical integration formulas in multiplicative calculus. Applying these results to the Quadrature formulas demonstrates their practical utility in numerical integration. Furthermore, numerical analysis provides empirical evidence of the effectiveness of the derived findings. It is also demonstrated that the newly proven inequalities extend certain existing results in the literature.

What's a parent to do? Measuring cultural logics of parenting with computational text analysis

Leading theories on parenting in the United States suggest that parenting varies widely by socioeconomic status, with middle-class parents practicing “concerted cultivation”—marked by parents' intensive efforts to foster their children's development—and working-class parents engaging in the “accomplishment of natural growth”—with children given more freedom to manage their own time. While frequently inferred that these parenting practices reflect different cultural logics of parenting, such logics are inherently hard to measure. Our paper proposes a new inductive way to study parenting logics using computational text analysis applied to a nationally representative survey where respondents provided parenting advice across three hypothetical parenting situations. Analyzing this advice using Biterm Topic Modeling we find that nearly all parenting logics reflect some form of intensive parenting, but within that are multiple nuanced versions varying across two dimensions: (1) assertive vs negotiated parenting, and (2) pedagogic vs pragmatic parenting. Using fractional multinomial logistic regression, we find little difference in how parenting logics vary by race/ethnicity, education, and income, suggesting more similarity across groups and more variability within groups than commonly understood. These findings also demonstrate how computational techniques may provide complementary tools to enrich the study of long-standing questions in social science research, at times offering an analytical naïveté that human coding cannot offer.

Automated Translation of Chronic Disease Diagnosis Codes using the ChatGPT Large Language Model

BackgroundThe International Classification of Diseases (ICD) is revised over time and there are region-specific versions, including ICD-10-CA (Canada) and ICD-9-CM (USA). Studies spanning multiple ICD versions require crosswalks to translate diagnosis codes across versions, but manual crosswalk development is costly and requires clinical expertise. ObjectiveTo evaluate the accuracy of a pre-trained large language model (LLM) to automatically translate chronic disease diagnosis codes from ICD-10-CA to ICD-9-CM. ApproachEight prompts were developed to instruct the OpenAI Generative Pre-trained Transformer 4 (GPT-4) LLM to translate 1,272 ICD-10-CA codes for the Elixhauser Comorbidity Index to ICD-9-CM. Prompt accuracy (%) was measured against a crosswalk developed by the Canadian Institute of Health Information. Variability was assessed by replicating each prompt three times. Mean accuracy ± standard deviation was reported for each prompt across replications, for both five-digit and truncated three-digit codes. ResultsThe highest prompt performance was observed when assigning a persona of a medical coding specialist (40.8% ± 0.9%), requesting justification for the selected code (41.4% ± 1.1%), and providing diagnosis code labels (47.5% ± 0.7%). For truncated three-digit codes, these prompts achieved accuracy of 82.0% ± 0.5%, 80.8% ± 0.9%, and 82.7% ± 0.1%, respectively. Combining these three prompting techniques marginally improved accuracy to 48.6% ± 0.7% for five-digit codes and 84.3% ± 0.2% for truncated three-digit codes. ConclusionGeneral-purpose LLMs are currently not sufficiently accurate at automating ICD code translation for chronic diseases. ImplicationsAdditional experiments with fine-tuning, task-specific training, and prompt engineering are needed to improve accuracy and reduce variability.