China

Multilingual Question Answering (QA) systems must ensure factual consistency across languages, especially for objective queries such as What is jaundice?, while also accounting for cultural variation in subjective responses. We propose MIND, a user-in-the-loop fact-checking pipeline to detect factual and cultural discrepancies in multilingual QA knowledge bases. MIND highlights divergent answers to culturally sensitive questions (e.g., Who assists in childbirth?) that vary by region and context. We evaluate MIND on a bilingual QA system in the maternal and infant health domain and release a dataset of bilingual questions annotated for factual and cultural inconsistencies. We further test MIND on datasets from other domains to assess generalization. In all cases, MIND reliably identifies inconsistencies, supporting the development of more culturally aware and factually consistent QA systems.

EMNLP 2025

Discrepancy Detection at the Data Level: Toward Consistent Multilingual Question Answering

multilingual qa

multilingual evaluation

knowledge base qa

human-in-the-loop

topic modeling

stages 

COLING, the International Conference on Computational Linguistics, is one of the premier conferences for natural language processing and computational linguistics. Often grouped within the field of artificial intelligence, but actually pre-dating the development of artificial intelligence, advances in computational linguistics and natural language processing are now some of the major drivers behind the use of artificial intelligence for commercial and social applications – for example, on-line search, machine translation and with voice-assisted conversational devices.

First established in 1965, the biennial COLING conference is held in diverse parts of the globe and attracts participants from both top-ranked research centers and emerging countries. Today, the most important developments in our field are taking place not only in universities and academic research institutes, but also in industrial research departments and in technological startups. COLING conferences provide opportunities for all these communities to showcase their exciting developments.

COLING 2020

COLING, the International Conference on Computational Linguistics, is one of the premier conferences for natural language processing and computational linguistics. Often grouped within the field of artificial intelligence, but actually pre-dating the development of artificial intelligence, advances in computational linguistics and natural language processing are now some of the major drivers behind the use of artificial intelligence for commercial and social applications – for example, on-line search, machine translation and with voice-assisted conversational devices.

technical 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."<br>

*Message from the General Chair, Dirk Hovy*





<br>

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

Dialectal data are characterized by linguistic variation that appears small to humans but has a significant impact on the performance of models. This dialect gap has been related to various factors (e.g., data size, economic and social factors) whose impact, however, turns out to be inconsistent. In this work, we investigate factors impacting the model performance more directly: we correlate Tokenization Parity (TP) and Information Parity (IP), as measures of representational biases in pre-trained multilingual models, with the downstream performance. We compare state-of-the-art decoder-only LLMs with encoder-based models across three tasks: dialect classification, topic classification, and extractive question answering, controlling for varying scripts (Latin vs. non-Latin) and resource availability (high vs. low). Our analysis reveals that TP is a better predictor of the performance on tasks reliant on syntactic and morphological cues (e.g., extractive QA), while IP better predicts performance in semantic tasks (e.g., topic classification). Complementary analyses, including tokenizer behavior, vocabulary coverage, and qualitative insights, reveal that the language support claims of LLMs often might mask deeper mismatches at the script or token level.

Tokenization and Representation Biases in Multilingual Models on Dialectal NLP Tasks

The human-centered word association test (WAT) serves as a cognitive proxy, revealing sociocultural variations through lexical-semantic patterns. We extend this test into an LLM-adaptive, free-relation task to assess the alignment of large language models (LLMs) with cross-cultural cognition. To mitigate the culture preference, we propose _CultureSteer_, an innovative prompt-free approach that integrates a culture-aware steering mechanism to guide semantic representations toward culturally specific spaces. Experiments show that current LLMs exhibit significant bias toward Western cultural (notably in American) schemas at the word association level. In contrast, our model substantially improves cross-cultural alignment, surpassing prompt-based methods in capturing diverse semantic associations. Further validation on culture-sensitive downstream tasks confirms its efficacy in fostering cognitive alignment across cultures. This work contributes a novel methodological paradigm for enhancing cultural awareness in LLMs, advancing the development of more inclusive language technologies.

From Word to World: Evaluate and Mitigate Culture Bias in LLMs via Word Association Test

Despite extensive efforts in safety alignment, large language models (LLMs) remain vulnerable to jailbreak attacks. Activation steering offers a training-free defense method but relies on fixed steering coefficients, resulting in suboptimal protection and increased false rejections of benign inputs. To address this, we propose AdaSteer, an adaptive activation steering method that dynamically adjusts model behavior based on input characteristics. We identify two key properties: Rejection Law (R-Law), which shows that stronger steering is needed for jailbreak inputs opposing the rejection direction, and Harmfulness Law (H-Law), which differentiates adversarial and benign inputs. AdaSteer steers input representations along both the Rejection Direction (RD) and Harmfulness Direction (HD), with adaptive coefficients learned via logistic regression, ensuring robust jailbreak defense while preserving benign input handling. Experiments on LLaMA-3.1, Gemma-2, and Qwen2.5 show that AdaSteer outperforms baseline methods across multiple jailbreak attacks with minimal impact on utility. Our results highlight the potential of interpretable model internals for real-time, flexible safety enforcement in LLMs.

AdaSteer: Your Aligned LLM is Inherently an Adaptive Jailbreak Defender

Enabling digital humans to express rich emotions has significant applications in dialogue systems, gaming, and other interactive scenarios. While recent advances in talking head synthesis have achieved impressive results in lip synchronization, they tend to overlook the rich and dynamic nature of facial expressions. To fill this critical gap, we introduce an end-to-end text-to-expression model that explicitly focuses on emotional dynamics. Our model learns expressive facial variations in a continuous latent space and generates expressions that are diverse, fluid, and emotionally coherent. To support this task, we introduce EmoAva, a large-scale and high-quality dataset containing 15,000 text–3D expression pairs. Extensive experiments on both existing datasets and EmoAva demonstrate that our method significantly outperforms baselines across multiple evaluation metrics, marking a significant advancement in the field.

When Words Smile: Generating Diverse Emotional Facial Expressions from Text

Existing video benchmarks often resemble image-based benchmarks, with questions like "What color is the woman's dress in this video?" or "What's the weather like in the video?" where models can answer using just a single frame or text, limiting the assessment of large vision language models (LVLMs) in terms of video perception and temporal reasoning abilities. To address this, we introduce GLIMPSE, a benchmark specifically designed to evaluate whether LVLMs truly understand video content. GLIMPSE includes 3,269 videos and over 4,342 highly visual-centric questions across 11 categories, such as Trajectory, Temporal QA, Forensics Detection, and Video Sorting, all meticulously curated and crafted by human annotators. Many of these questions require watching the entire video and reasoning based on video context to answer accurately, as they cannot be addressed using only a few frames or text information. Human testing on GLIMPSE achieves an average accuracy of 94.82%, but GLIMPSE poses significant challenges for existing LVLMs. Even the best-performing model, Gemini 1.5 Pro, reaches only 56.98% accuracy, indicating that current LVLMs still have a long way to go in truly understanding and interpreting video content.

 GLIMPSE: Do Large Vision-Language Models Truly Think With Videos or Just Glimpse at Them?

With the continued proliferation of Large Language Model (LLM) based chatbots, there is a growing demand for generating responses that are not only linguistically fluent but also consistently aligned with persona-specific traits in conversations. However, existing role-play and persona-based chat approaches rely heavily on static role descriptions, coarse-grained signal space, and low-quality synthetic data, which fail to capture dynamic fine-grained details in human-like chat. Human-like chat requires modeling subtle latent traits, such as emotional tone, situational awareness, and evolving personality, which are difficult to predefine and cannot be easily learned from synthetic or distillation-based data. To address these limitations, we propose a Verbal Variational Auto-Encoding (V-VAE) framework, containing a variational auto-encoding module and fine-grained control space which dynamically adapts dialogue behaviour based on fine-grained, interpretable latent variables across talking style, interaction patterns, and personal attributes. We also construct a high-quality dataset, HumanChatData, and benchmark HumanChatBench to address the scarcity of high-quality data in the human-like domain. Experiments show that LLMs based on V-VAE consistently outperform standard baselines on HumanChatBench and DialogBench, which further demonstrates the effectiveness of V-VAE and HumanChatData.

V-VAE: A Variational Auto Encoding Framework Towards Fine-Grained Control over Human-Like Chat

Can small language models (0.5B–5B parameters) meaningfully engage in trauma-informed, empathetic dialogue for individuals with PTSD? We answer this by introducing TIDE, a dataset of 10,000 two-turn dialogues across 500 diverse PTSD client personas, grounded in a three-factor empathy model: emotion recognition, distress normalization, and supportive reflection. All scenarios and reference responses were reviewed for realism and trauma sensitivity by a clinical psychologist specializing in PTSD. Eight small language models are evaluated before and after fine-tuning, with outputs compared to a frontier model (Claude Sonnet 3.5) as reference. Our IRB-approved human evaluation and automatic metrics reveal that, while fine-tuning generally improves perceived empathy, gains are highly scenario- and user-dependent, with smaller models facing an "empathy ceiling." Notably, demographic analyses show older adults value distress validation and graduate-educated users prefer nuanced replies, while gender effects are minimal. We highlight limitations of automatic metrics and the need for context- and user-aware system design. Our findings—along with the planned release of TIDE—offer a foundation for building safe, resource-efficient, and ethically sound empathetic AI to supplement, not replace, clinical mental health care.

The Pursuit of Empathy: Evaluating Small Language Models for PTSD Dialogue Support

Relational databases are central to modern data management, yet most data exists in unstructured forms like text documents. To bridge this gap, we leverage large language models (LLMs) to automatically synthesize a relational database by generating its schema and populating its tables from raw text. We introduce SQUiD, a novel neurosymbolic framework that decomposes this task into four stages, each with specialized techniques. Our experiments show that SQUiD consistently outperforms baselines across diverse datasets.

SQUiD: Synthesizing Relational Databases from Unstructured Text

Large language models (LLMs) excel at complex reasoning tasks but often suffer from overconfidence and computational inefficiency due to fixed computation budgets and miscalibrated confidence estimates. We present a novel framework for computationally efficient, trustworthy reasoning under uncertainty, introducing two complementary techniques: Diversity-Aware Self-Signal Dilution (DASD) and Convergent Adaptive Weighted Sampling (CAWS). DASD operates in an unsupervised manner to dilute overconfident, semantically redundant reasoning paths, thereby producing better-calibrated internal confidence estimates. CAWS dynamically allocates computational resources at inference time by aggregating these signals and terminating computation once answer dominance and stability are achieved. Comprehensive experiments across three reasoning datasets demonstrate that our approach maintains accuracy levels while achieving over 70\% reduction in inference cost, surpassing competitive baselines. Our framework provides a scalable, unsupervised solution for reliable and efficient LLM reasoning.

Reasoning under Uncertainty: Efficient LLM Inference via Unsupervised Confidence Dilution and Convergent Adaptive Sampling

Large language models segment many words into multiple tokens, and there is mixed evidence as to whether tokenization affects how state-of-the-art models represent meanings. Chinese characters present an opportunity to investigate this issue: They contain semantic radicals, which often convey useful information; characters with the same semantic radical tend to begin with the same one or two bytes (when using UTF-8 encodings); and tokens are common strings of bytes, so characters with the same radical often begin with the same token. This study asked GPT-4, GPT-4o, and Llama 3 whether characters contain the same semantic radical, elicited semantic similarity ratings, and conducted odd-one-out tasks (i.e., which character is not like the others). In all cases, misalignment between tokens and radicals systematically corrupted representations of Chinese characters. In experiments comparing characters represented by single tokens to multi-token characters, the models were less accurate for single-token characters, which suggests that segmenting words into fewer, longer tokens obscures valuable information in word form and will not resolve the problems introduced by tokenization. In experiments with twelve European languages, misalignment between tokens and suffixes systematically corrupted categorization of words by all three models, which suggests that the tendency to treat malformed tokens like linguistic units is pervasive.

Downloads

Next from EMNLP 2025

Tokenization and Representation Biases in Multilingual Models on Dialectal NLP Tasks

Similar lecture

Handling Anomalies of Synthetic Questions in Unsupervised Question Answering

Stay up to date with the latest Underline news!

PRESENTATIONS

CONFERENCES

COMPANY

RESOURCES