United States

This paper presents our submission to Multi-
ClinAI’s NER subtask for #SMM4H-HeaRD
2026. We focus on the questions 1) which
Language Model represents the clinical notes
best and 2) which annotations can help train-
ing these models. To get answers for these
questions, we follow a token-based classifica-
tion approach with pretrained encoder language
models, where we compare models that were
pretrained on generic data against medical data,
and on a single language, Dutch, against many
languages. In addition, we present two data-
augmented systems: one with data from the
other languages of the workshop for multilin-
gual training, and one with synthetic annota-
tions1
.

ACL 2026

LotusOrchid at #SMM4H–HeaRD 2026: Fitting pretrained encoders for
Dutch medical data

token classification

domain adaptation

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. <br>
[Full Message from General Chair Philipp Koehn](https://drive.google.com/file/d/1QiJFOmMjvDgsDo3i1XzlQnImxe3RU_Cy/view?usp=sharing) <br><br><br>
<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>
<br><br><br><br>

To access this event page, you need to log in with the **email address you registered with**. <br>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.

workshop paper

This paper presents our approach to Task 5 of
the #SMM4H-HeaRD 2026 Workshop, which
focuses on detecting patient metadata in SARS-
CoV-2 sequencing articles as a binary classi-
fication task. We explore both encoder-based
and large language model (LLM) approaches,
using BioM-BERT as a baseline and Mistral-
Nemo as the LLM. To improve performance,
we propose a paraphrase-based data augmenta-
tion pipeline using Qwen3, where paraphrased
training and validation instances are added for
fine-tuning. For the LLM, we perform prompt
refinement and error analysis, while for the
encoder-based model, we reformulate the task
as a hypothesis-conditioned classification task
inspired by Natural Language Inference (NLI).
Our methods improve both models: Mistral-
Nemo increases from 0.423 to 0.750 F1, and
BioM-BERT from 0.801 to 0.821 on the valida-
tion set. Although Mistral-Nemo does not sur-
pass BioM-BERT, our best BioM-BERT model
achieves an F1-score of 0.786 on the test set,
outperforming the mean and median of com-
peting systems. To support reproducibility, we
release our best-performing model on Hugging
Face.

PEI at #SMM4H-HeaRD 2026: Enhancing Patient Metadata Detection via
Hypothesis-Conditioned Classification and Paraphrase-Based Data
Augmentation

We present a system for multilingual clini-
cal named entity recognition (NER) submitted
to the MultiClinNER subtask of MultiClinAI
2026, covering all seven languages and three
entity classes (disease, symptom, procedure).
Our approach trains one binary token classifier
ensemble per entity class using cross-lingual
fine-tuning of XLM-RoBERTa-large, with all
languages handled jointly. We apply character-
level ensembling over six models (two encoder
variants × three cross-validation folds). This
ensembling method provides more granular
probability estimates than single-model clas-
sifiers, allowing for more flexible precision-
recall trade-off tuning. The system achieves
character-level F1 scores of 0.70–0.88 on the
official test set.

Discovery@FI at #SMM4H–HeaRD 2026:
Ensemble Character Classifier for Multilingual Clinical NER

In this paper, we describe systems for two
#SMM4H-HeaRD 2026 shared tasks. Task 6
asks for per-axis TNM cancer staging from
free-text TCGA pathology reports under se-
vere label imbalance and long-document con-
straints. We fine-tune GatorTron-base sep-
arately on each axis using Focal loss with
class weights and a pooled [CLS]–mean rep-
resentation, reaching macro F1 of 0.700 (T),
0.774 (N), and 0.640 (M) on test set 2 against
a baseline of 0.454, 0.591, and 0.554 re-
spectively. Task 7 asks for span-level detec-
tion of opioid-related ClinicalImpacts and So-
cialImpacts in first-person Reddit posts. We
combine DeBERTa-large and PubMedBERT
(two seeds each) in a uniform-weight ensemble
with boundary-aware loss, entity-replacement
augmentation, and a first-person post filter,
achieving strict F1 of 0.51 and relaxed F1 of
0.60, above both the task mean (0.46 / 0.55) and
median (0.48 / 0.58).

CUET_DiagNLP at #SMM4H-HeaRD 2026: Per-Axis TNM Staging from
Pathology Reports and Opioid Impact Span Detection from Social Media

We present an overview of the MultiClinAI shared task, which focuses on multilingual clinical entity extraction and automatic corpus generation through annotation projection. It addresses two key challenges in clinical natural language processing (NLP): (i) developing comparable multilingual named entity recognition (NER) systems and (ii) automatically constructing multilingual clinical corpora through annotation projection. The MultiClinAI task provides a unified benchmark for evaluating multilingual and cross-lingual clinical NLP approaches that cover diseases, symptoms, and procedures in Spanish, English, Dutch, Italian, Romanian, Swedish, and Czech. A total of 21 teams from 13 countries participated, submitting 531 runs across the different subtasks. The top runs obtained very competitive results across several languages and entity types. The results highlight both the challenges and opportunities of multilingual clinical information extraction. All resources, including a corpus of over 738,201 manually revised entity mentions across seven languages, are publicly available on Zenodo at: https://zenodo.org/records/19334278.

The MultiClinAI Shared Task on Multilingual Clinical Corpus Construction and Concept Extraction: Systems, Evaluation, and Datasets

We present two systems for the #SMM4H-
HeaRD 2026 Task 2 (Lopez-Garcia et al., 2026)
shared task of automated insomnia detection
from clinical notes. Our system addresses both
subtasks: (1) binary insomnia classification and
(2) multi-label rule prediction with evidence
span extraction. For Subtask 1, we employ an
ensemble architecture combining Qwen3-4B-
Instruct1 and Bio_ClinicalBERT2 to capture
both general semantic reasoning and domain-
specific clinical representations. The frame-
work utilizes chunk-based processing with over-
lapping token windows to handle long clinical
notes efficiently. For Subtask 2, we develop
a dual-head multi-task transformer model that
jointly predicts insomnia labels and token-level
evidence spans using BIO tagging3. To im-
prove clinical relevance, we additionally incor-
porate sentence-level filtering using sentence-
transformer embeddings and similarity-based
retrieval of insomnia-related contexts. Experi-
mental results suggest competitive performance
relative to the shared task mean and median
scores across both subtasks. Our best Subtask 1
system achieves a recall of 0.9474, surpassing
the shared-task mean and median recall, while
our Subtask 2 system exceeds the mean and
median scores across label classification, exact
match, and partial match metrics. The end-
to-end implementation is publicly available on
GitHub.

NoviceTrio in #SMM4H-HeaRD 2026: Hybrid Clinical Transformer
Ensembles for Insomnia Detection and Evidence Extraction from Clinical
Notes

The Tumor, Node, and Metastasis (TNM) stag-
ing system is critical to cancer treatment. This
study aims to predict TNM stage labels in-
dependently, with the Cancer Genome Atlas
(TCGA) pathology report as the sixth shared
task of SMM4H-HeaRD 2026. The problem is
framed as three multi-label classification tasks.
We explore both classical and deep learning ap-
proaches using Term Frequency-Inverse Doc-
ument Frequency (TF-IDF) features and em-
beddings from ClinicalBERT, BioBERT, and
PubMedBERT. These representations are used
with Logistic Regression (LR), Light Gradient
Boosting Machine (LightGBM), Feed-Forward
Neural Networks (FFNN), and Wide Residual
Networks (WRN). Our results show that in-
dividual embeddings perform similarly to the
TNM label classification, while their combi-
nation improves its predictive ability. WRN
achieves AUROC scores of 0.839 (T), 0.8502
(N), and 0.803 (M) with F1-scores of 0.622,
0.702, and 0.9337, respectively, for the train-
ing phase. LightGBM with TF-IDF performs
best with AUROC scores of 0.9368 (T), 0.9524
(N), and 0.8311 (M) and F1-scores of 0.7559
(T), 0.7384 (N), and 0.7017 (M) during the
training phase. Furthermore, the result of the
Codabench for the test sets indicates a Macro-
F1 score of 0.978, 0.957, and 0.879 for the T,
N, and M categories respectively for test set
1; while test set 2 records a Macro-F1 score
for T, N, and M is 0.807, 0.767, 1.0 respec-
tively. However, performance declined during
the evaluation phase of the test sets, a drop from
0.938 for test set 1 to 0.858 for test set 2, for the
Macro-F1 score across all stages; suggesting
limitations in model generalizability, sensitivity
to class imbalance, and challenges in process-
ing lengthy clinical documents. Although this
study provides an efficient baseline model and
a reproducible pipeline, further optimization
and validation are required before it can be con-
sidered suitable for use in a real-world clinical
setting.

CaresAI at SMM4H-HeaRD 2026: Predicting TNM Staging

In this paper, we address Task 1 on adverse
drug event (ADE) detection and Task 8 on Mul-
tiClinNER at SMM4H-HeaRD 2026. ADE
detection is formulated as a multilingual bi-
nary classification problem over social me-
dia posts spanning German, French, Russian,
English, Mandarin and Japanese, with zero-
shot on Farsi. Using XLM-RoBERTa-Large
with a dual-pooling head, combined with strat-
ified sampling, language-conditioned inputs,
translation-based augmentation, and calibrated
ensembling, our model achieves a macro F1
score of 0.6088, surpassing both the competi-
tion mean (0.5465) and median (0.5798). Our
work in MultiClinNER targets clinical NER
for English text. Using GLiNER-large with
sliding-window inference, query augmentation,
and calibrated thresholds, it achieves strict F1
scores of 0.7591 (Disease), 0.7263 (Procedure),
and 0.6733 (Symptom), outperforming a Pub-
MedBERT baseline across all entities.

Vinland_Vector at #SMM4H-HeaRD 2026: Multilingual ADE Detection
and Query-Augmented Clinical NER for English

We present our systems for Task 1 and Task 3
of the #SMM4H-HeaRD 2026 shared tasks.
Task 1 focuses on binary classification of ad-
verse drug event (ADE) mentions across seven
languages, including a zero-shot Persian set-
ting without labeled training data. We fine-tune
XLM-RoBERTa-large using weighted cross-
entropy loss and augment low-resource settings
with additional CADEC data and machine-
translation-based Persian augmentation. Our
system achieves a macro F1 score of 0.582, out-
performing the shared task average of 0.547.
Task 3 addresses influenza vaccine effective-
ness estimation through classification of vac-
cination status and flu-test results from X
posts. We fine-tune twitter-roberta-large,
achieving micro F1 scores of 0.845 for vaccina-
tion status and 0.883 for flu-test classification
on the official test set. Post-evaluation experi-
ments with focal loss, test-time augmentation,
and head-tail truncation further improve per-
formance. These results highlight the effec-
tiveness of robust transformer adaptation for
health-related social media classification.

Cuet_Data_Wizards at #SMM4H-HeaRD 2026:
Multilingual ADE Detection and Influenza Vaccine Effectiveness
Estimation from Social Media

We introduce Output-Space Search (OS-Search), which turns LLM generation into endpoint search. An outer loop selects a target z* in a frozen encoder-defined 3D output space Z, and a retrieval-grounded policy trained with sequence-level RL generates outputs whose coordinates land near z* under standard autoregressive decoding. This enables parallel sweeps and black-box optimization in Z without path-dependent token/program search. On stories, sweeping Z (text) yields 3.1x higher LLM-scored diversity than prompt-chaining. On code, Bayesian optimization over Z (code) improves an objective withheld from the controller under matched inference budgets while preserving validity.

Output-Space Search: Targeting LLM Generations in a Frozen Encoder-Defined Output Space

Code generation is increasingly critical for real-world applications. Still, diffusion-based large language models continue to struggle with this demand. Unlike free-form text, code requires syntactic precision; even minor structural inconsistencies can render a program non-executable. Existing diffusion-based large language models rely on random token masking for corruption, leading to two key failures: they lack awareness of syntactic boundaries during the iterative denoising process, and they fail to capture the long-range hierarchical dependencies essential for program correctness.

We propose TreeDiff to address both issues. Specifically, we propose a syntax-aware diffusion framework that incorporates structural priors from Abstract Syntax Tree (AST) into the corruption process. Instead of masking individual tokens at random, we selectively mask tokens belonging to key AST nodes. By aligning the corruption process with the underlying structure of code, our method encourages the model to internalize the compositional nature of programming languages, enabling it to reconstruct programs that respect grammatical boundaries and capture long-range dependencies. Our method achieves a 13.3% relative improvement over the random masking training method, demonstrating its effectiveness in code generation task by leveraging underlying structures.

Downloads

Next from ACL 2026

PEI at #SMM4H-HeaRD 2026: Enhancing Patient Metadata Detection via Hypothesis-Conditioned Classification and Paraphrase-Based Data Augmentation

Similar lecture

MedMind AI at #SMM4H-HeaRD 2026: Data Extraction and Generation Using Prompt Engineering and Structured Outputs (Tasks 1–6)

Stay up to date with the latest Underline news!

PRESENTATIONS

CONFERENCES

COMPANY

RESOURCES

.css-70qvj9{display:-webkit-box;display:-webkit-flex;display:-ms-flexbox;display:flex;-webkit-align-items:center;-webkit-box-align:center;-ms-flex-align:center;align-items:center;}Downloads