Morocco

Distilling advanced Large Language Models&#39; instruction-following capabilities into smaller models using a selected subset has become a mainstream approach in model training. While existing synthetic instruction data selection strategies rely mainly on single-dimensional signals (i.e., reward scores, model perplexity), they fail to capture the complexity of instruction-following across diverse fields. Therefore, we investigate more diverse signals to capture comprehensive instruction-response pair characteristics and propose three foundational metrics that leverage Multi-LLMs wisdom, informed by (1) diverse LLM responses and (2) reward model assessment. Building upon base metrics, we propose CROWDSELECT, an integrated metric incorporating a clustering-based approach to maintain response diversity. Our comprehensive experiments demonstrate that our foundation metrics consistently improve performance across 4 base models on MT-bench and Arena-Hard. CROWDSELECT, efficiently incorporating all metrics, achieves state-of-the-art performance in both Full and LoRA fine-tuning, showing improvements of 4.81% on Arena-Hard and 11.1% on MT-bench with Llama-3.2-3b-instruct. We hope our findings will bring valuable insights for future research in this direction.

EACL 2026 Main Conference

CROWDSELECT: SyntheticInstruction Data Selection with Multi-LLM Wisdom

multi-llm

distillation

Distilling advanced Large Language Models' instruction-following capabilities into smaller models using a selected subset has become a mainstream approach in model training. While existing synthetic instruction data selection strategies rely mainly on single-dimensional signals (i.e., reward scores, model perplexity), they fail to capture the complexity of instruction-following across diverse fields. Therefore, we investigate more diverse signals to capture comprehensive instruction-response pair characteristics and propose three foundational metrics that leverage Multi-LLMs wisdom, informed by (1) diverse LLM responses and (2) reward model assessment. Building upon base metrics, we propose CROWDSELECT, an integrated metric incorporating a clustering-based approach to maintain response diversity. Our comprehensive experiments demonstrate that our foundation metrics consistently improve performance across 4 base models on MT-bench and Arena-Hard. CROWDSELECT, efficiently incorporating all metrics, achieves state-of-the-art performance in both Full and LoRA fine-tuning, showing improvements of 4.81% on Arena-Hard and 11.1% on MT-bench with Llama-3.2-3b-instruct. We hope our findings will bring valuable insights for future research in this direction.

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)*<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://underline.io/events/522/reception','_blank')">Go to Workshops and Tutorials Program</button></html>
<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 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. 

We conduct a systematic audit of three widely used reasoning benchmarks, SocialIQa, FauxPas-EAI, and ToMi, and uncover pervasive flaws in both benchmark items and evaluation methodology. Using five LLMs (GPT-{3, 3.5, 4, o1}, and LLaMA 3.1) as diagnostic tools, we identify structural, semantic, and pragmatic issues in benchmark design (e.g., duplicated items, ambiguous wording, and implausible answers), as well as scoring procedures that prioritize output form over reasoning process. Through systematic human annotation and re-evaluation on cleaned benchmark subsets, we find that model scores often improve not due to due to erratic surface wording variations and not to improved reasoning. Infact, further analyses show that model performance is highly sensitive to minor input variations such as context availability and phrasing, revealing that high scores may reflect alignment with format-specific cues rather than consistent inference based on the input. These findings challenge the validity of current benchmark-based claims about reasoning in LLMs, and highlight the need for evaluation protocols that assess reasoning as a process of drawing inference from available information, rather than as static output selection. We release audited data and evaluation tools to support more interpretable and diagnostic assessments of model reasoning.

Garbage In, Reasoning Out? Why Benchmark Scores are Unreliable and What to Do About It

Robustly evaluating the long-form storytelling capabilities of Large Language Models (LLMs) remains a significant challenge, as existing benchmarks often lack the necessary scale, diversity, or objective measures. To address this, we introduce WebNovelBench, a novel benchmark specifically designed for evaluating long-form novel generation. WebNovelBench leverages a large-scale dataset of over 4,000 Chinese web novels, framing evaluation as a synopsis-to-story generation task. We propose a multi-faceted framework encompassing eight narrative quality dimensions, assessed automatically via an LLM-as-Judge approach. Scores are aggregated using Principal Component Analysis and mapped to a percentile rank against human-authored works. Our experiments demonstrate that WebNovelBench effectively differentiates between human-written masterpieces, popular web novels, and LLM-generated content. We provide a comprehensive analysis of 24 state-of-the-art LLMs, ranking their storytelling abilities and offering insights for future development. This benchmark provides a scalable, replicable, and data-driven methodology for assessing and advancing LLM-driven narrative generation.

WebNovelBench: Placing LLM Novelists on the Web Novel Distribution

The rapid rise of deepfake technology poses a severe threat to social and political stability by enabling hyper-realistic synthetic media capable of manipulating public perception. However, existing detection methods struggle with two core limitations: (1) modality fragmentation, which leads to poor generalization across diverse and adversarial deepfake modalities; and (2) shallow inter-modal reasoning, resulting in limited detection of fine-grained semantic inconsistencies. To address these, we propose ConLLM (Contrastive Learning with Large Language Models), a hybrid framework for robust multimodal deepfake detection. ConLLM employs a two-stage architecture: stage 1 uses Pre-Trained Models (PTMs) to extract modality-specific embeddings; stage 2 aligns these embeddings via contrastive learning to mitigate modality fragmentation, and refines them using LLM-based reasoning to address shallow inter-modal reasoning by capturing semantic inconsistencies. ConLLM demonstrates strong performance across audio, video, and audio-visual modalities. It reduces audio deepfake EER by up to 50%, improves video accuracy by up to 8%, and achieves approximately 9% accuracy gains in audio-visual tasks. Ablation studies confirm that PTM-based embeddings contribute 9%вЂ“10% consistent improvements across modalities. Our code and data is available at: https://anonymous.4open.science/r/ConLLM-0D8A/README.md

Revealing the Truth with ConLLM for Detecting Multi-Modal Deepfakes

Large Language Model (LLM) safety is inherently pluralistic, reflecting variations in moral norms, cultural expectations, and demographic contexts. Yet, existing alignment datasets such as Anthropic-HH and DICES rely on demographically narrow annotator pools, overlooking variation in safety perception across communities. Demo-SafetyBench addresses this gap by modeling demographic pluralism directly at the prompt level, decoupling value framing from responses. In Stage I, prompts from DICES are reclassified into 14 safety domains (adapted from BeaverTails) using Mistral-7B-Instruct-v0.3, retaining demographic metadata and expanding low-resource domains via Llama-3.1-8B-Instruct with SimHash-based deduplication, yielding 43,050 samples. In Stage II, pluralistic sensitivity is evaluated using LLMs-as-Raters—Gemma-7B, GPT-4o, and LLaMA-2-7B—under zero-shot inference. Balanced thresholds (delta = 0.5, tau = 10) achieve high reliability (ICC = 0.87) and low demographic sensitivity (DS = 0.12), confirming that pluralistic safety evaluation can be both scalable and demographically robust. Code and data available at: https://github.com/usmaann/Demo-SafetyBench

Do Large Language Models Reflect Demographic Pluralism in Safety?

We design, implement, and evaluate adversarial decoding, a new, generic text generation technique that produces readable documents for adversarial objectives such as RAG poisoning, jailbreaking, and evasion of defensive filters. Prior generation methods either produce easily detectable gibberish (even methods that optimize for low perplexity), or cannot handle objectives that include embedding similarity. In particular, they cannot produce readable adversarial documents that (1) are retrieved by RAG systems in response to broad classes of queries, and (2) adversarially influence subsequent generation. We measure the effectiveness of adversarial decoding for different objectives and demonstrate that it outperforms existing methods while producing adversarial documents that cannot be automatically distinguished from natural documents by fluency and readability.

Adversarial Decoding: Generating Readable Documents for Adversarial Objectives

Large Language Models (LLMs) have been frequently used as automatic annotators for tasks such as Text Emotion Recognition (TER). We consider a scenario in which annotators assign at least one emotion label from a large set of options to a text snippet. For this emotion tagging task, we propose a novel zero-shot algorithm that leverages Best-Worst Scaling (BWS), prompting the LLM to choose the least and most suitable emotions for a given text from several label subsets. The LLM's choices can be represented by a graph linking labels via worse-than relations. Random walks on this graph yield the final score for each label. We compare our algorithm with naive prompting approaches as well as an established BWS-based method. Extensive experiments demonstrate the suitability of the method. It proves to compare favorably to the benchmarks in terms of both accuracy and calibration with respect to human annotations. Moreover, our algorithm's automatic annotations are shown to be suitable for finetuning lightweight emotion classification models. The proposed method consumes considerably fewer computational resources than the established BWS approach.

Training-Free Text Emotion Tagging via LLM-Based Best-Worst Scaling

Text-to-SQL systems translate natural language questions into executable SQL queries, and recent progress with large language models (LLMs) has driven substantial improvements in this task. Schema linking remains a critical component in Text-to-SQL systems, reducing prompt size for models with narrow context windows and sharpening model focus even when the entire schema fits. We present a zero-shot, training-free schema linking approach that first constructs a schema graph based on foreign key relations, then uses a single prompt to a lightweight LLM to extract source and destination tables from the user query, followed by applying classical path-finding algorithms and post-processing to identify the optimal sequence of tables and columns that should be joined, enabling the LLM to generate more accurate SQL queries. To handle real-world databases where foreign keys may be missing or inconsistent, we further propose an LLM-guided joinability discovery step that infers table connections before graph construction, ensuring robustness across diverse schemas. Despite being simple, cost-effective, and highly scalable, our method achieves state-of-the-art results on both the BIRD and Spider 2.0 benchmarks, outperforming previous specialized, fine-tuned, and complex multi-step LLM-based approaches.

SchemaGraphSQL: Efficient Schema Linking with Pathfinding Graph Algorithms for Text-to-SQL on Large-Scale Databases

Although expressive TTS systems aim to capture human-like emotion, little is known about how well emotional signals in text correspond to those in speech. In this short paper, we investigate the emotion alignment (Valence, Arousal, Dominance) between text and speech models. We use 8 large language models for identifying text emotions and two audio models for speech analysis, across three genres: Podcasts, Audiobooks and TED talks. Findings show that while language models perform well on emotion recognition from situational text, and the audio models performs well on speech, they show alignment on the Valence dimension only. Further, the genre of the content significantly impacts the alignment performance: audiobooks exhibit better text-audio correlation than TED talks. Our results highlight the limitations of text-based emotion expression in TTS systems: performance depends on genre, is mostly limited to Valence, and more context fails to improve emotional alignment between text and speech.

Emotion Alignment Between Text and Speech is Limited: A Cross-Modal Study

The significance of tasks entrusted to LLM-based research assistants (agents) and the associated societal risks are increasing each year. Agents are being explored in critical domains such as medicine, finance, law, infrastructure, and other sensitive applications that require system transparency and high user trust. The quality of these agents is typically evaluated by accuracy, sometimes extended to partial correctness. In this position paper, we argue that this focus on outcomes is insufficient as it can obscure risky agent behaviours such as skipping important steps, hallucinating tool use, relying on outdated parametric knowledge and other means of bypassing recommended processes. Our core position is that a holistic agent evaluation must include process evaluation, especially for critical applications. We conduct a small-scale study to assess the feasibility of automatic process evaluation, present a proof-of-concept compliance score, analyse use cases of bad and good behaviours, and offer recommendations for a more holistic evaluation of agents.

Process Evaluation for Agentic Systems

Professionals working in technical domain typically hand-draw (on whitebpoard, paper, etc.) technical diagrams (e.g., flowcharts, block diagrams, etc.) during discussions, however, later if they want to edit these, it needs to be drawn from scratch. Modern day VLMs have made tremendous progress in image understanding but these struggle when it comes to understanding technical diagrams. One way to overcome this problem is to fine-tune on real world hand-drawn images, but it is not practically possible to generate large number of such images. In this resource paper, we introduce a large synthetically generated corpus (reflective of real world images) for training VLMs and subsequently evaluate VLMs on a smaller corpus of hand-drawn images (with the help of humans). We introduce several new self-supervision tasks for training and perform extensive experiments with various baseline models and fine-tune Llama 3.2 11B-instruct model on synthetic images on these tasks to obtain LLama-VL-TUG, which significantly improves the ROUGE-L performance of Llama 3.2 11B-instruct by 2.14x and achieves best all round performance across all baseline models. On real-world images, human evaluation reveals that we achieve minimum compilation errors across all baselines in 7 out of 8 diagram types and improve average F1 score of Llama 3.2 11B-instruct by 6.97x.

Downloads

Next from EACL 2026 Main Conference

Garbage In, Reasoning Out? Why Benchmark Scores are Unreliable and What to Do About It

Stay up to date with the latest Underline news!

PRESENTATIONS

CONFERENCES

COMPANY

RESOURCES