China

Access to high-quality labeled data remains a limiting factor in applied supervised learning. Active learning (AL), a popular approach to optimizing the use of limited annotation budgets in training ML models, often relies on at least one of several simplifying assumptions, which rarely hold in practice when acknowledging human label variation (HLV). Label variation (LV), i.e., differing labels for the same instance, is common, especially in natural language processing. Yet annotation frameworks often still rest on the assumption of a single ground truth, overlooking HLV, i.e., the occurrence of plausible differences in annotations, as an informative signal. 
In this paper, we examine foundational assumptions about truth and label nature, highlighting the need to decompose observed LV into signal (e.g., HLV) and noise (e.g., annotation error). We survey how the AL and (H)LV communities have addressed---or neglected---these distinctions and propose a conceptual framework for incorporating HLV throughout the AL loop, including instance selection, annotator choice, and label representation. We further discuss the integration of large language models (LLM) as annotators. Our work aims to lay a conceptual foundation for (H)LV-aware active learning, better reflecting the complexities of real-world annotation.

EMNLP 2025

Revisiting Active Learning under (Human) Label Variation

philosophy of annotation

annotator selection function

human label variation

acquisition function

supervised learning

active learning

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

ACL 2026

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

## Welcome!
"I am excited to welcome you to this year’s edition of the Conference on Empirical Methods in Natural Language Processing! Importantly, it marks the 30th edition of EMNLP. With over 8,000 submissions, more than 3,000 accepted papers, and thousands of attendees, we have come a long way from that first
workshop, which had 14 accepted papers. As the field looks ahead, Suzhou is the fitting location for celebrating this milestone: rooted in a long literary tradition, yet modern and forward-looking, and home to a large share of the NLP community." 

*Message from the General Chair, Dirk Hovy*

Celebrate 30 Years of EMNLP! 
EMNLP 2025 will be held in Suzhou, China from November 5th to November 9th, 2025.

For datasets to accurately represent diverse opinions in a population, they must preserve variation in data labels while filtering out spam or low-quality responses. How can we balance annotator reliability and representation? We empirically evaluate how a range of heuristics for annotator filtering affect the preservation of variation on subjective tasks. We find that these methods, designed for contexts in which variation from a single ground-truth label is considered noise, often remove annotators who disagree instead of spam annotators, introducing suboptimal tradeoffs between accuracy and label diversity. We find that conservative settings for annotator removal (<5%) are best, after which all tested methods increase the mean absolute error from the true average label. We analyze performance on synthetic spam to observe that these methods often assume spam annotators are more random than real spammers tend to be: most spammers are distributionally indistinguishable from real annotators, and the minority that are distinguishable tend to give fixed answers, not random ones. Thus, tasks requiring the preservation of variation reverse the intuition of existing spam filtering methods: spammers tend to be less random than non-spammers, so metrics that assume variation is spam fare worse. These results highlight the need for spam removal methods that account for label diversity.

Balancing Quality and Variation: Spam Filtering Distorts Data Label Distributions

This position paper argues that annotation disagreement in Natural Language Inference (NLI) is not mere noise but often reflects meaningful variation, especially when triggered by ambiguity in the premise or hypothesis. While underspecified guidelines and annotator behavior contribute to variation, content-based ambiguity provides a process-independent signal of divergent human perspectives. We call for a shift toward ambiguity-aware NLI that first identifies ambiguous input pairs, classifies their types, and only then proceeds to inference. To support this shift, we present a framework that incorporates ambiguity detection and classification prior to inference. We also introduce a unified taxonomy that synthesizes existing taxonomies, illustrates key subtypes with examples, and motivates targeted detection methods that better align models with human interpretation. Although current resources lack datasets explicitly annotated for ambiguity and subtypes, this gap presents an opportunity: by developing new annotated resources and exploring unsupervised approaches to ambiguity detection, we enable more robust, explainable, and human-aligned NLI systems.

From Disagreement to Understanding: The Case for Ambiguity Detection in NLI

Irony is a subjective and pragmatically complex phenomenon, often conveyed through rhetorical figures and interpreted differently across individuals. In this study, we adopt a perspectivist approach, accounting for the socio-demographic background of annotators, to investigate whether specific rhetorical strategies promote a shared perception of irony within demographic groups, and whether Large Language Models (LLMs) reflect specific perspectives.
Focusing on the Italian subset of the perspectivist MultiPICo dataset, we manually annotate rhetorical figures in ironic replies using a linguistically grounded taxonomy. The annotation is carried out by expert annotators balanced by generation and gender, enabling us to analyze inter-group agreement and polarization. Our results show that some rhetorical figures lead to higher levels of agreement, suggesting that certain rhetorical strategies are more effective in promoting a shared perception of irony.
We fine-tune multilingual LLMs for rhetorical figure classification, and evaluate whether their outputs align with different demographic perspectives. Results reveal that models show varying degrees of alignment with specific groups, reflecting potential perspectivist behavior in model predictions.
These findings highlight the role of rhetorical figures in structuring irony perception and underscore the importance of socio-demographics in both annotation and model evaluation.

Towards a Perspectivist Understanding of Irony through Rhetorical Figures

Toxicity labels at sub-document granularity and disaggregated labels lead to more nuanced and personalized toxicity classification and facilitate analysis. We re-annotate a subset of 1983 posts of the Jigsaw Toxic Comment Classification Challenge and provide disaggregated toxicity labels and spans that identify inappropriate language and targets of toxic statements.

Manual analysis shows that five annotations per instance effectively capture meaningful disagreement patterns and allow for finer distinctions between genuine disagreement and that arising from annotation error or inconsistency. Our main findings are: (1) Disagreement often stems from divergent interpretations of edge-case toxicity. (2) Disagreement is especially high in cases of toxic statements involving non-human targets. (3) Disagreement on whether a passage consists of inappropriate language occurs not only on inherently questionable terms, but also on words that may be inappropriate in specific contexts while remaining acceptable in others. (4) Transformer-based models effectively learn from aggregated data that reduces false negative classifications by being more sensitive towards minority opinions for posts to be toxic. We publish the new annotations under the CC BY 4.0 license.

A Disaggregated Dataset on English Offensiveness Containing Spans

Plain Language Summaries (PLS) play a critical role in improving health literacy, enabling informed decision-making and equitable healthcare access. However, writing PLS requires domain expertise and is time-consuming, making automation a valuable strategy for improving accessibility at scale. Automated methods often prioritize efficiency over comprehension, and the unique simplification requirements of medical documents challenge generic solutions. We present a multi-agent system for generating PLS, using Cochrane PLS as a proof of concept. The system decomposes simplification in four tasks, each handled by specialized agents: information extraction, writing, diagnostic, and evaluation. It integrates a medical glossary (20,637 terms) and a statistical analyzer that evaluates text patterns to guide revisions. We evaluated on 100 Cochrane abstracts using three models: Gemini-2.5-Pro, GPT-5 and the open model GPT-OSS-120B. The system achieved superior performance across semantic similarity, factual alignment, and readability metrics compared to single-prompt baselines. By combining AI agents with specific evaluation tools, this work offers a scalable solution that reduces the health literacy gap by making medical information more understandable to the public through accurate, readable summaries.

A Multi-Agent Framework with Diagnostic Feedback for Iterative Plain Language Summary Generation from Cochrane Medical Abstracts

Individuals with intellectual disabilities often have difficulties in comprehending complex texts. While many text-to-image models prioritize photorealism over cognitive accessibility, it is not clear how visual illustrations relate to text simplifications (TS) generated from them. This paper presents a structured vision language model (VLM) prompting framework for generating cognitively accessible images from simplified texts. We designed five prompt templates, i.e., Basic Object Focus, Contextual Scene, Educational Layout, Multi-Level Detail, and Grid Layout, each following distinct spatial arrangements while adhering to accessibility constraints such as object count limits, spatial separation, and content restrictions. Using 400 sentence-level TS pairs from four established text simplification datasets (OneStopEnglish, SimPA, Wikipedia, ASSET), we conducted a two-phase evaluation: Phase 1 assessed template effectiveness with CLIP similarity scores, and Phase 2 involved expert annotation of generated images across ten visual styles by four accessibility specialists. Results show that the Basic Object Focus template achieved the highest semantic alignment, indicating that visual minimalism enhances accessibility. Expert evaluation further identified Retro style as the most accessible and Wikipedia as the most effective text source. Inter-annotator agreement varied across dimensions, with Text Simplicity showing strong reliability and Image Quality proving more subjective. Overall, our framework offers practical guidelines for accessible content creation and underscores the importance of structured prompting in AI-generated visual accessibility tools.

Template-Based Text-to-Image Alignment for Language Accessibility: A Study on Visualizing Text Simplifications

Automatic readability assessment plays a key role in ensuring effective communication between humans and language models. Despite significant progress, the field is hindered by inconsistent definitions of readability and measurements that rely on surface-level text properties. In this work, we investigate the factors shaping human perceptions of readability through the analysis of 1.2k judgments, finding that, beyond surface-level cues, information content and topic strongly shape text comprehensibility. Furthermore, we evaluate 15 popular readability metrics across 5 datasets, contrasting them with 5 more nuanced, model-based metrics. Our results show that four model-based metrics consistently place among the top 4 in rank correlations with human judgments, while the best performing traditional metric achieves an average rank of 7.8. These findings highlight a mismatch between current readability metrics and human perceptions, pointing to model-based approaches as a more promising direction.

Readability Reconsidered: A Cross-Dataset Analysis of Reference-Free Metrics

This work presents a system for on-device text simplification that enables users to process sensitive text without relying on cloud-based services. Through the use of quantization techniques and a novel approach to controllable text simplification, we reduce model size by up to 75% with minimal performance degradation. Our models demonstrate efficient state-of-the-art results using a synthetic dataset of 2,909 examples, outperforming prior work trained on 300K examples. This efficiency stems from: (1) a single control token strategy that precisely targets specific reading levels, (2) a contrastive training approach that enriches model understanding through exposure to multiple simplification levels, and (3) individual models that dedicate full parameter capacity to specific reading level transformations. Our best models achieve up to 82.18 BLEU (at the Advanced level) and 46.12 SARI (at the Elementary level) on standard benchmarks, with performance preserved even after aggressive quantization. This work is implemented as a collaboration with the Mozilla AI team to processes text entirely locally, ensuring sensitive information never leaves the user's device. We have a demonstration video (https://youtu.be/TzmaxnARMzg) and a web demo available at: (https://pablorom2004.github.io/Simplification-Web-Demo)

Efficient On-Device Text Simplification for Firefox with Synthetic Data Fine-Tuning

In this contribution to the CEFR level simplification TSAR 2025 Shared Task, we propose two systems, EZ-SCALAR and SAGA, that implement two differing approaches to prompting LLMs for proficiency-adapted simplification. Our results place us in the middle of the participating teams, and reveal that using external lexical resources to guide simplification improves overall results.

GRIPF at TSAR 2025 Shared Task: Towards controlled CEFR level simplification with the help of inter-model interactions

Automatic Text Simplification (TS) makes complex texts more accessible but often lacks control over target readability levels. We propose a lightweight, prompt-based approach to English TS that explicitly aligns outputs with CEFR proficiency standards. Our method employs a three-stage pipeline, guided by rule-informed prompts inspired by expert strategies. In the TSAR 2025 Shared Task, our system achieved competitive performance, with stronger results at B1 level and challenges at A2 level due to over-simplification. These findings highlight the promise of prompt-based CEFR-oriented simplification and the need for more flexible constraint design.

Downloads

Next from EMNLP 2025

Balancing Quality and Variation: Spam Filtering Distorts Data Label Distributions

Similar lecture

Position: From Noise to Signal to Selbstzweck - Reframing Human Label Variation in the Era of Post-training in NLP

.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

Next from EMNLP 2025

Balancing Quality and Variation: Spam Filtering Distorts Data Label Distributions

Similar lecture

Position: From Noise to Signal to Selbstzweck - Reframing Human Label Variation in the Era of Post-training in NLP

Downloads