Morocco

Large language models (LLMs) have demonstrated impressive few-shot in-context learning (ICL) abilities. Still, we show that they are sometimes prone to a `copying bias&#39;, where they copy answers from provided examples instead of learning the underlying patterns. In this work, we propose a novel and simple method to mitigate such copying bias. First, we create a synthetic task and use the Integrated Gradients method to identify neurons that prioritize copying over generalization. We demonstrate that pruning these neurons consistently improves performance across a diverse set of ICL tasks, including both single-token and multi-token scenarios, while maintaining or even improving the model&#39;s general capabilities. We also show that our method is applicable across various LLM architectures, including Transformers and State-Space Models, without requiring modifications. In our analysis, we adopt a task-recognition perspective on ICL and examine task vectors (Hendel et al., 2023) induced by the model. We find that pruning enhances the quality of these vectors, suggesting that the pruned neurons previously hindered effective task recognition.

EACL 2026

Mitigating Copy Bias in In-Context Learning through Neuron Pruning

in-context-learning

bias

**Welcome to EMNLP 2023!**

On behalf of the EMNLP 2023 Organizing Committee, I extend a warm and heartfelt welcome to all of you to the 2023 Conference on Empirical Methods in Natural Language Processing (EMNLP). It is with immense pleasure and excitement that we gather here in Singapore, a vibrant hub of innovation and technological advancement.

The conference program is packed with insightful presentations, thought-provoking workshops, and engaging networking opportunities. In addition to the technical sessions, we have also planned several social events that will provide you with the opportunity to connect with your colleagues.

The conference is held in person at the Resorts World Convention Centre in Singapore, and available on-line with the help of Underline.

For those in Singapore, I hope that you will find time to explore the exciting Sentosa Island, Gardens by the Bay, Singapore Botanic Gardens and the many other attractions unique to Singapore.

Of course, we are grateful to our sponsors and partners for their generous support of EMNLP 2023, whose contributions make it possible for us to host this world-class event.

Yuji Matsumoto (RIKEN AIP) 
EMNLP 2023 General Chair

To access the **EMNLP 2023** event page on Underline, you need to register for the Conference. 
Please follow **[this link](https://2023.emnlp.org/registration/)** for more details.

EMNLP 2023

EMNLP 2023 took place in Singapore from Dec 6th to Dec 10th, 2023.

Large language models (LLMs) have demonstrated impressive few-shot in-context learning (ICL) abilities. Still, we show that they are sometimes prone to a `copying bias', where they copy answers from provided examples instead of learning the underlying patterns. In this work, we propose a novel and simple method to mitigate such copying bias. First, we create a synthetic task and use the Integrated Gradients method to identify neurons that prioritize copying over generalization. We demonstrate that pruning these neurons consistently improves performance across a diverse set of ICL tasks, including both single-token and multi-token scenarios, while maintaining or even improving the model's general capabilities. We also show that our method is applicable across various LLM architectures, including Transformers and State-Space Models, without requiring modifications. In our analysis, we adopt a task-recognition perspective on ICL and examine task vectors (Hendel et al., 2023) induced by the model. We find that pruning enhances the quality of these vectors, suggesting that the pruned neurons previously hindered effective task recognition.

technical paper

#### *Message from the General Chair, Aline Villavicencio*
I’m delighted and honoured to welcome you to the 19th Conference of the European Chapter of the Association for Computational Linguistics (EACL 2026), taking place in the beautiful city of Rabat, in Morocco, in March 24-29, 2026. EACL is the flagship European conference of the Association and EACL 2026 proudly continues our field’s tradition of excellence in scholarship, innovation, and inclusivity. I am deeply grateful to the many volunteers whose dedication, generosity, and tireless efforts have made this conference possible.
For the first time EACL is being hosted in the African continent. This is an important milestone for our community, and we are grateful to our Moroccan hosts for enabling this historic moment by bringing this edition of EACL to Rabat. We are also delighted that the Second Arabic NLP School is co-located with EACL. We hope attendees enjoy this wonderful opportunity to strengthen ties with the Computational Linguistics communities across the African continent. *[Read full message](https://drive.google.com/file/d/14NlmHvwM6fPJuMmOvVh7K0vtQbEyv3SZ/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://underline.io/events/522/reception','_blank')">Go to Workshops and Tutorials Program</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 EACL 2026". Please be sure to check your spam email folder if you do not see an email confirmation right away.

Please log in

To access this event page, you are required to register.
Please complete your registration to continue.

We recommend reading [**the registration information**](https://2026.eacl.org/registration/) first.

**Online Registration Form**: https://acl.swoogo.com/eacl2026

Registration Required

Welcome to the 19th Conference of the European Chapter of the Association for Computational Linguistics (EACL). EACL 2026 will be held in Rabat, Morocco, from March 24–29, 2026. 

Large Language Models (LLMs) exhibit varying levels of Natural Language Understanding (NLU), with larger models performing better but at higher computational costs. To optimize cost-effectiveness, simpler queries should be handled by smaller, more efficient models, while complex tasks are escalated to larger ones. We propose a confidence-driven strategy that dynamically selects the most suitable model based on confidence estimates. By assessing a model's confidence in handling the task and response accuracy, tasks are delegated to a larger model when necessary, ensuring reliability while minimizing computation. Experiments on the Massive Multitask Language Understanding (MMLU) benchmark show that our approach achieves accuracy comparable to the largest model while reducing computational costs by 20% to 40%. When applied to GPT-4o API calls, it reduces token usage by approximately 60%, further improving cost efficiency. These findings underscore the potential of confidence-based model selection to enhance real-world LLM deployment, particularly in resource-constrained settings such as edge devices and commercial API applications.

Confidence-Driven Multi-Scale Model Selection for Cost-Efficient Inference

Large Language Models (LLMs) have achieved remarkable progress, yet their reliance on parametric knowledge often leads to hallucinations. Retrieval-Augmented Generation (RAG) mitigates this issue by grounding outputs in external documents, where the quality of retrieval is critical. While retrieval methods have been widely benchmarked in English, it remains unclear which approaches are most effective for Vietnamese, a language characterized by informal queries, noisy documents, and limited resources. Prior studies are restricted to clean datasets or narrow domains, leaving fragmented insights. To the best of our knowledge, we present the first comprehensive benchmark of retrieval methods for Vietnamese across multiple real-world domains. We systematically compare lexical, dense, and hybrid methods on datasets spanning education, legal, healthcare, customer support, lifestyle, and Wikipedia, and introduce two new datasets capturing authentic educational counseling and customer service interactions. Beyond reporting benchmark numbers, we distill a set of empirical insights that clarify trade-offs, highlight domain-specific challenges, and provide practical guidance for building robust Vietnamese QA systems. Together, these contributions offer the first large-scale, practice-oriented perspective on Vietnamese retrieval and inform both academic research and real-world deployment in low-resource languages. All datasets and evaluation scripts are available atВ https://anonymous.4open.science/r/ViRE.

Which Works Best for Vietnamese? A Practical Study of Information Retrieval Methods across Domains

Humanitarian Mine Action (HMA) addresses the challenge of detecting and removing landmines from conflict regions. Much of the life-saving operational knowledge produced by HMA agencies is buried in unstructured reports, limiting the transferability of information between agencies. To address this issue, we propose TextMine: the first dataset, evaluation framework and ontology-guided large language model (LLM) pipeline for knowledge extraction in the HMA domain. TextMine structures HMA reports into (subject, relation, object)-triples, thus creating domain-specific knowledge. To ensure real-world relevance, we created the dataset in collaboration with Cambodian Mine Action Center (CMAC). We further introduce a bias-aware evaluation framework that combines human-annotated triples with an LLM-as-Judge protocol to mitigate position bias in reference-free scoring. Our experiments show that ontology-aligned prompts improve extraction accuracy by up to 44.2%, reduce hallucinations by 22.5%, and enhance format adherence by 20.9% compared to baseline models. We publicly release the dataset and code.

TextMine: Data, Evaluation Framework and Ontology-guided LLM Pipeline for Humanitarian Mine Action

In recent years, pre-trained large language models (LLMs) have become a cornerstone for automatically generating answers in question-and-answer (Q&A) communities, significantly reducing user wait times and improving response quality. However, these models require substantial computational resources and are prone to generating hallucinated or unreliable content. To overcome these limitations, we propose an advanced expert-oriented Retrieval-Augmented Generation (RAG) framework as a cost-effective and reliable alternative. Central to our approach is a user-aware question entailment recognition module, which leverages user modeling to identify archived questions with answers that fully or partially address the user's new query. This user modeling significantly improves retrieval relevance, resulting in reduced hallucination and enhanced answer quality. The framework synthesizes expert-written answers from similar questions to generate unified responses. Experimental results on the CQADupStack and SE-PQA datasets show the superiority of our user-aware approach over its user-agnostic counterpart, with ROUGE-1 gains of 3.6% and 0.9%. Both human and AI evaluations confirm the effectiveness of incorporating user modeling in minimizing hallucination and delivering contextually appropriate answers, demonstrating its potential for real-world Q&A systems. The code and data are available on a GitHub repository at https://anonymous.4open.science/r/User-Oriented-RAG-CQA.

Enhancing Reliability in Community Question Answering with an Expert-Oriented RAG System

Sequential test-time scaling is a promising training-free method to improve large reasoning model accuracy, but as currently implemented, significant limitations have been observed. Inducing models to think for longer can increase their accuracy, but as the length of reasoning is further extended, it has also been shown to result in accuracy degradation and model instability. This work presents a novel sequential test-time scaling method, Min-Seek, which improves model accuracy significantly, over a wide range of induced thoughts, stabilizing the accuracy of sequential scaling, and removing the need for reasoning length fine-tuning. Beyond improving model accuracy over a variety of reasoning tasks, our method is inherently efficient, as only the KV pairs of one additional induced thought are kept in the KV cache during reasoning. With a custom KV cache which stores keys without position embeddings, by dynamically encoding them contiguously before each new generated thought, our method can continue to reason well beyond a model's maximum context length, and under mild conditions has linear computational complexity.

Thinking Long, but Short: Stable Sequential Test-Time Scaling for Large Reasoning Models

The rapid growth of academic video content makes it difficult for students and educators to find relevant information efficiently. This is especially challenging for low-resource languages like Bengali due to the lack of a video summarization tool. This paper presents the first end-to-end pipeline for the abstractive summarization of Bengali academic videos. The proposed system preprocesses audio to improve transcription quality and converts speech to text using Google’s Speech Recognition API. The text is segmented using a smart chunking method to be compatible with the model's context window. For summarization, we fine-tuned the BanglaT5 model on a new benchmark dataset of 10,029 text-summary pairs obtained from educational videos. To generate relevant titles, we fine-tuned the mT5-multilingual-XLSum model on our curated dataset of 1,005 summary-title pairs. Our fine-tuned summarization model shows strong performance, achieving F1 scores of 0.8793 (BERTScore), 0.3894 (ROUGE-1), and 0.2557 (ROUGE-L), outperforming other models. Our title generation model achieved ROUGE-1 and ROUGE-L F1 scores of 0.4476 and 0.3720, respectively. The summaries include timestamps for easy video navigation. This work aims to improve the accessibility of educational content in Bengali. It also contributes valuable datasets and a robust baseline system that demonstrates strong zero-shot capabilities on other spoken contents. The code and datasets are available at https://github.com/LamisaDeya/Bangla-Video-Summarization.

Abstractive Summarization of Bengali Academic Videos Based on Audio Subtitles

Recent work has explored the use of personal information in the form of persona sentences or self-disclosures to improve modeling of individual characteristics and prediction of annotator labels for subjective tasks. The volume of personal information has historically been restricted and thus little exploration has gone into understanding what kind of information is most informative for predicting annotator labels. In this work, we categorize self-disclosure sentences and use them to build annotator models for predicting judgments of social norms. We perform several ablations and analyses to examine the impact of the type of information on our ability to predict annotation patterns. We find that demographics are more impactful than attitudes, relationships, and experiences. Generally, theory-based approaches worked better than automatic clusters. Contrary to previous work, only a small number of related comments are needed. Lastly, having a more diverse sample of annotator self-disclosures leads to the best performance.

Examining the Utility of Self-disclosure Types for Modeling Annotators of Social Norms

Despite significant ongoing efforts in safety alignment, large language models (LLMs) such as GPT-4 and LLaMA 3 remain vulnerable to jailbreak attacks that can induce harmful behaviors, including through the use of adversarial suffixes. Building on prior research, we hypothesize that these adversarial suffixes are not mere bugs but may represent features that can dominate the LLM's behavior. To evaluate this hypothesis, we conduct several experiments. First, we demonstrate that benign features can be effectively made to function as adversarial suffixes, i.e., we develop a feature extraction method to extract sample-agnostic features from benign dataset in the form of suffixes and show that these suffixes may effectively compromise safety alignment. Second, we show that adversarial suffixes generated from jailbreak attacks may contain meaningful features, i.e., appending the same suffix to different prompts results in responses exhibiting specific characteristics. Third, we show that such benign-yet-safety-compromising features can be easily introduced through fine-tuning using only benign datasets. As a result, we are able to completely eliminate GPT's safety alignment in a blackbox setting through finetuning with only benign data. Our code and data is available at \url{anonymous.4open.science/r/suffix-maybe-features-D17C/}.

Unleashing the Unseen: Harnessing Benign Datasets for Jailbreaking Large Language Models

Large language models (LLMs) have demonstrated remarkable performance in abstractive summarization tasks. However, their ability to precisely control summary attributes (e.g., length or topic) remains underexplored, limiting their adaptability to specific user preferences. In this paper, we systematically explore the controllability of LLMs. To this end, we revisit summary attribute measurements and introduce iterative evaluation metrics, failure rate and average iteration count, to more precisely evaluate controllability beyond assessment of errors. Our findings show that LLMs struggle more with numerical attributes than with linguistic attributes. To address this challenge, we propose a guide-to-explain framework (GTE) for controllable summarization. GTE enables the model to identify misaligned attributes in the initial draft and guides it to self-explain errors in the previous output. By encouraging reflection on attribute misalignment, GTE generates well-adjusted summaries that satisfy the desired attributes with robust effectiveness while requiring surprisingly fewer iterations than other iterative approaches.

Exploring Iterative Controllable Summarization with Large Language Models

Evaluating visual activity recognition systems is challenging due to inherent ambiguities in verb semantics and image interpretation. When describing actions in images, synonymous verbs can refer to the same event (e.g., *brushing* vs. *grooming*), while different perspectives can lead to equally valid but distinct verb choices (e.g., *piloting* vs. *operating*). Standard exact-match evaluation, which relies on a single gold answer, fails to capture these ambiguities, resulting in an incomplete assessment of model performance. To address this, we propose a vision-language clustering framework that constructs **verb sense clusters**, providing a more robust evaluation. Our analysis of the imSitu dataset shows that each image maps to an average of 2.8 sense clusters, with each cluster representing a distinct perspective of the image. We evaluate multiple activity recognition models and compare our cluster-based evaluation with standard evaluation methods. Additionally, our human alignment analysis suggests that the cluster-based evaluation better aligns with human judgements, offering a more nuanced assessment of model performance.

Downloads

Next from EACL 2026

Confidence-Driven Multi-Scale Model Selection for Cost-Efficient Inference

Similar lecture

Estimating Large Language Model Capabilities without Labeled Test Data