United States

Large language models (LLMs) can generate or synthesize clinical text for a wide range of applications, from improving clinical documentation to augmenting clinical text analytics. Yet evaluations typically focus on a narrow aspect -- such as similarity or utility comparisons -- even though these aspects are complementary and best viewed in parallel. In this study, we aim to conduct a systematic evaluation of LLM-generated clinical text, which includes intrinsic, extrinsic, and factuality evaluations of synthetic clinical notes rephrased from MIMIC databases at million-note scale. Our analysis demonstrates that synthetic notes preserve core clinical information and predictive utility for coarse-grained tasks despite substantial linguistic changes, but lose fine-grained details for task like ICD coding. We show this loss of detail can be substantially mitigated by rephrasing notes by chunks rather than by the whole note, but at the cost of reduced factual precision under incomplete context. Through fact-checking and error analysis, we further find that synthesis errors are dominated by misinterpretation of clinical context, alongside temporal confusion, measurement errors, and fabricated claims. Finally, we show that the synthetic notes -- despite their task-agnostic nature -- can effectively augment task-specific training for rare ICD codes.

ACL 2026

Systematic Evaluation of the Quality of Synthetic Clinical Notes Rephrased by LLMs at Million-Note Scale

rephrasing

synthetic data

fact-checking

clinical nlp

evaluation

### Welcome to ACL 2023, the 61st Annual Meeting of the Association for Computational Linguistics! 
 The conference will be held in Toronto, Canada, July 9-14, 2023. 
Following the succession of the recent conferences in our field, ACL 2023 will adopt a hybrid format.
While the impact of Covid has considerably diminished in terms of traveling, obtaining visas to Canada
entails a very long process. Moreover, the global economic conditions pose challenges for many individuals to travel to conferences. Recognizing these circumstances, we know many participants may not be
able to attend the conference in person. Therefore, we are committed to providing a great virtual platform
so everyone has the opportunity to interact with other participants and enjoy the conference. Based on the
current registered participants, approxiately 30% have chosen to attend the conference virtually. Whether
you join us in person or virtually, we sincerely hope everyone has a remarkable conference experience. 
This General Chair’s message is where I express my gratitude to the many individuals who have made
enormous contributions to the conference over the past year.

Read [**ACL 2023 General Chair's message**](https://docs.google.com/document/d/1WobYM7norbG4dI48s75HfJoD89qgX5a_F-6U8AteLSA/edit?usp=sharing/) in full.

##### **[Conference Handbook](https://2023.aclweb.org/downloads/acl2023-handbook.pdf)**

ACL 2023

The Association for Computational Linguistics (ACL) is the premier international scientific and professional society for people working on computational problems involving human language, a field often referred to as either computational linguistics or natural language processing.

workshop paper

It is my great pleasure and honor to welcome you to the 64th Annual Meeting of the Association for Computational Linguistics (ACL 2026), held in San Diego, California, USA, July 2–7, 2026. ACL 2026 continues our fields tradition of excellence in scholarship, innovation, and inclusivity, and I am deeply grateful to the many volunteers who have worked tirelessly to bring this event to life.

It is exciting to see the many developments in the field reflected at our conference, such as the advancement of large language models (born out of our community) into many application directions. This is not without challenges, partly due to unsolved problems, and partly due to misuse of the technology. All this is reflected well at ACL 2026. The conference also reflects that the field is not only about advancing technology, but also the science of fundamentally understanding language and how it is use by many different speaker communities. 
[Full Message from General Chair Philipp Koehn](https://drive.google.com/file/d/1QiJFOmMjvDgsDo3i1XzlQnImxe3RU_Cy/view?usp=sharing) 
<html><button style="display: inline-flex; align-items: center; justify-content: center; white-space: nowrap; border-radius: 9999px; font-weight: bold; background: #7c3aed; color: white; font-family: 'Space Grotesk', sans-serif; height: 40px; font-size: 16px; padding: 0 20px; border: none; cursor: pointer" onclick="window.open('https://drive.google.com/file/d/1QiJFOmMjvDgsDo3i1XzlQnImxe3RU_Cy/view?usp=sharing')">Conference Handbook</button></html>

To access this event page, you need to log in with the **email address you registered with**. Access credentials will be sent to your email from Underline - subject line "Welcome to ACL 2026". Please be sure to check your spam email folder if you do not see an email confirmation right away.

To access this event page, you are required to register.
Please complete your registration to continue.

We recommend reading [**the registration information**](https://2026.aclweb.org/registration/) first.

Registration Required

The 64th Annual Meeting of the Association for Computational Linguistics (ACL 2026) will take place in San Diego, California, from July 2 - 7, 2026.

Clinical coding maps clinical documentation to standardized medical codes, an essential yet time-consuming administrative task that could benefit from automation. Current models on ICD coding are typically optimized for codes from a specific ICD version. However, in reality, ICD systems evolve continuously, and different versions are adopted across time periods and regions. Moreover, ICD coding suffers from the long-tail problem, and rare code performance can be a bottleneck for developing implementable models. We examine whether it is viable to train version-independent models by combining data annotated in different ICD versions, which may help address these challenges. We add ICD-9 data to the training of a modified label-wise attention model for ICD-10 prediction, and find that despite the version mismatch, adding ICD-9 yields a 27% increase in micro F1 for 18K rare ICD codes compared to training on ICD-10 alone. On 8K frequent ICD-10 codes, the multi-version training also substantially improves macro metrics, with far fewer model parameters.

Bridging the Version Gap: Multi-version Training Improves ICD Code Prediction, Especially for Rare Codes

The advent of single-cell RNA sequencing has enabled unprecedented resolution of cell fate decisions and regulatory mechanisms during peri-implantation human embryogenesis, in which accurate cell type annotation is a fundamental prerequisite and the first step for subsequent fate and mechanism inference. Large language models (LLMs) have demonstrated outstanding performance in various fields. However, current studies mostly rely on traditional methods and have not explored the application of LLMs in the field of human embryonic cell annotation. The main reason is the lack of instruction tuning datasets and evaluation benchmarks. In this paper, we proposed EmCellLLM, the first open sourced LLMs that are specialized for human embryonic cell type prediction task based on fine-tuning Qwen3-8B with EmCell4Instruction, the first embryonic cell type prediction instruction dataset. To support LLM instruction tuning, we also build EmCellBench, the first benchmark for evaluating human embryonic cell type prediction ability of LLMs. We compare our models with a variety of LLMs on EmCellBench, where our model outperforms all other open-sourced LLMs as well as DeepSeek.

EmCellLLM: Human Peri-Implantation Embryonic Cell Annotation Based on Large Language Models

Large Language Models (LLMs) are no longer mere laboratory objects of study. LLMs have become everyday tools in society across diverse populations and domains. In clinical contexts, LLMs have already been devised as clinical support applications. However, along with benefits, negative or adverse effects might arise, such as LLMs potentially providing psychologically distressing advice to adolescents when used for mental health support. This raises questions on the benefits of LLMs and calls for real-world evaluations: Are LLMs really helpful and effective for the intended purposes 
 people are using them or will use them for? To answer this type of question we propose to use Randomized Controlled Trials (RCTs). RCTs are considered the most strict experimental design in the fields of Medicine, Psychiatry, Psychology, among others; however, the use of RCTs in the NLP field is almost negligible. In spite of the NLP field being the de facto locus of research on LLMs, other fields, prominently Medicine, are leading the RCT evaluations on LLMs. In this primer paper, we present a concise introduction to the principles of RCTs to guide NLP researchers to design RCT studies 
 for evaluating LLMs.

Randomized Controlled Trials as the Gold-Standard for Evaluating LLMs: A Primer for Biomedical NLP Researchers

While humans can easily produce various types of answers, such as definitions, examples or paraphrases, Large Language Models (LLMs) struggle to provide correct answers to medical questions that require diverse answer formats. In this paper, we introduce TrackList, a fine-grained linguistic and statistical analysis pipeline to investigate the impact of the pre-training data on LLMs answers to diverse linguistic queries. We also propose RefoMed-EN, a medical dataset consisting of 6,170 human-annotated medical terms alongside their corresponding definitions, denominations, exemplifications, explanations, or paraphrases. We investigated whether the high or low frequency of a concept (head or tail knowledge) impacts the language model's performance for answering medical questions. We evaluated the quality of the LLM's output using syntactic and semantic similarity metrics, statistical correlations and embeddings. Results showed that the LLM's answer quality for definition-type questions is the highest, while for the exemplification-type being the lowest. Additionally, we showed that for definition-type medical questions ("What is multiple sclerosis?"), LLMs are prone to paraphrase more for popular medical concepts, and less on more specialized medical knowledge.

TrackList: Tracing Back Query Linguistic Diversity for Head and Tail Medical Knowledge in Open Large Language Models

Discharge instructions are patient-facing, safety-critical documents that guide medication use, follow-up care, and recovery after hospitalization. Because they must synthesize information across the clinical record and often include post-discharge guidance not stated verbatim in the EHR, they are a difficult target for clinical text generation. In this work, we study discharge instructions in MIMIC-IV through a grounding-first lens. Using two LLMs, we decompose each discharge instruction into medically relevant statements and verify them against the Electronic Health Record (EHR). We find that discharge instructions for Surgical admissions are much longer, averaging roughly 24--25 statements per admission versus 11--12 in Non-Surgical cases, while supported content remains similar in absolute count. The additional Surgical content is dominated by statements that are not directly stated in the record or require clinically plausible extrapolation. Through this analysis, we advocate for better grounding and completeness evaluations at a fine-grained level, establishing a foundational step toward safer and more reliable discharge-instruction generation.

Discharge Instructions are not One Task: Grounding Differences Between Surgical and Non-Surgical Admissions

Prion diseases are rare, rapidly progressive, and fatal neurodegenerative disorders that remain difficult to diagnose, particularly in their early stages because of nonspecific clinical presentations. However, to our knowledge, there is no publicly available prion-disease-focused dataset designed to capture a broad range of clinically relevant entities from the biomedical literature. We introduce PrionNER, a manually annotated named entity recognition dataset for prion disease clinical information in PubMed abstracts. The current release comprises 317 abstracts, 2,943 sentences, and 6,955 text-bound entity annotations spanning 15 coarse-grained and 31 fine-grained clinically oriented entity types covering diseases, symptoms, diagnostics, findings, anatomy, treatments, and temporal and statistical evidence. Inter-annotator agreement reaches 81.78 exact-match F1, indicating strong annotation consistency. We benchmark supervised BERT baselines, W2NER, and zero-shot extractors on PrionNER. W2NER is the strongest supervised model, and Gemma-4-31B is the strongest zero-shot model, but the benchmark remains challenging, especially for structurally complex mentions and fine-grained clinically adjacent label distinctions. PrionNER provides a clinically grounded benchmark for prion-disease information extraction and supports research on rare-disease biomedical NLP under low-resource, fine-grained, and non-flat extraction conditions. The dataset, annotation guidelines, and evaluation scripts are available at https://github.com/daotuanan/PrionNER/

PrionNER: A Named Entity Recognition Dataset for Prion Disease Biomedical Literature

Individuals with particular mental health disorders may find it difficult to learn about their own condition. Therefore, efforts have been made to create materials that explain complex medical information in simpler words, which are also beneficial for caregivers and others. However, text simplification is commonly done in English and only sporadically in other languages. In this study, we explore potential ways for language-agnostic medical text simplification for the mental health domain. Our approach is to simplify the ICD-11 articles on primary psychotic disorders in English, German and French, using small LMs and various metrics for evaluating different aspects of the texts: lexical complexity and readability. Our results show that acceptable texts were produced only in English, and that a joint analysis of Measure of Textual Lexical Diversity (MTLD) and Flesch Reading Ease (FRE) provides the most insight, capturing both the best outcomes and signaling different types of issue. The study is preliminary and requires further investigation.

Evaluation of Multilingual Text Simplification for the Mental Health Domain: Exploring Small Language Models

Patient portal messages often embed clinical questions inside long, emotionally nuanced narratives, requiring clinicians to infer the underlying information need. We study the task of rewriting verbose patient-authored narratives into concise, clinician-interpreted questions framed as if querying an electronic health record (EHR) system. We evaluate a lightweight LLM-based rewrite pipeline that constrains outputs to 10-15 words and uses rule-based validation with regeneration. We test the approach on 140 distinct patient questions drawn from the ArchEHR-QA dataset and shared task. Each system output is double-annotated by two annotators for quality (Good/Ok/Bad) and error types (Generic, Malformed, Tangential, Hallucination). Results show that while models follow output constraints, they often produce overly generic or tangential questions, and occasional hallucinations introduce unsupported clinical details. Across both clinician-question and patient-narrative comparison settings, automatic metrics show substantial overlap across human quality labels; in pairwise meta-evaluation, BERTScore is the strongest proxy for human preferences. We release our code and annotations to support future work.

Reading Between the Lines: Toward Translating Verbose Patient-authored Messages into Clinician-Formulated Questions

Clinical documentation is essential for patient care, billing, and medical research, but it is subject to entrenched bias. While manual chart reviews can identify such bias, they are labor-intensive and expert-dependent. We introduce and evaluate StigMAD, a Multi-Agent Debate framework leveraging open-source Large Language Models (LLMs) to detect stigmatizing language in clinical documentation. We investigate reasoning (multi-agent debate), self-reflection, and self-consistency within this framework. Extensive experiments on clinical notes and patient summaries demonstrate that our framework provides significant advantages over rule-based and supervised baselines. A domain-specific LLM (MedGemma) achieved its highest performance using the StigMAD reasoning framework, while a general-purpose LLM (Llama) showed superior results with the self-consistency framework. These findings suggest that open-source LLMs, steered by structured prompting and reflective reasoning, can effectively support the scalable auditing of stigmatizing language, marking a critical step toward more equitable clinical NLP systems.

Investigating Stigmatizing Language in Clinical Documentation with Open-Source Large Language Models

Recent high-complexity agentic systems such as DeepRare perform strongly on rare disease diagnosis benchmarks, but it remains unclear when gains come from structured knowledge access and when they come from parametric
 LLM knowledge. We compare phenotypebased retrieval, LLM reranking, and unrestricted LLM diagnosis across seven benchmarks covering 10,382 cases. We find a clear performance crossover driven by retrieval coverage—the fraction of cases whose true diagnosis is within the retriever's top-50: on highcoverage datasets, ontology-based retrieval dominates; on low-coverage datasets, openended LLM diagnosis takes the lead. Building on this, adding an LLM reranker over retrieved candidates further improves accuracy across our patient-case benchmarks, closing most of the remaining gap to agentic systems (within 2 pp on MME and LIRICAL). We trace the crossover to two structural failure modes
 of ontology-based retrieval—annotation sparsity and phenotypic homogeneity—and show that aggregate scores across mixed benchmarks can hide these qualitatively different diagnostic settings. These findings motivate per-dataset evaluation and hybrid diagnostic systems that combine retrieval, reranking, and parametric LLM generation based on case characteristics.

Downloads

Next from ACL 2026

Bridging the Version Gap: Multi-version Training Improves ICD Code Prediction, Especially for Rare Codes

Similar lecture

Making the Most Out of the Limited Context Length: Predictive Power Varies with Clinical Note Type and Note Section