China

As large language models (LLMs) are used in complex writing workflows, users engage in multi-turn interactions to steer generations to better fit their needs. Rather than passively accepting output, users actively refine, explore, and co-construct text. We conduct a large-scale analysis of this collaborative behavior for users engaged in writing tasks in the wild with two popular AI assistants, Bing Copilot and WildChat. Our analysis goes beyond simple task classification or satisfaction estimation common in prior work and instead characterizes how users interact with LLMs through the course of a session. We identify prototypical behaviors in how users interact with LLMs in prompts following their original request. We refer to these as Prototypical Human-AI Collaboration Behaviors (PATHs) and find that a small group of PATHs explain a majority of the variation seen in user-LLM interaction. These PATHs span users revising intents, exploring texts, posing questions, adjusting style or injecting new content. Next, we find statistically significant correlations between specific writing intents and PATHs, revealing how users&#39; intents shape their collaboration behaviors. We conclude by discussing the implications of our findings on LLM alignment.

EMNLP 2025

Prototypical Human-AI Collaboration Behaviors from LLM-Assisted Writing in the Wild

interactive and collaborative generation

human-ai interaction

human behavior analysis

As large language models (LLMs) are used in complex writing workflows, users engage in multi-turn interactions to steer generations to better fit their needs. Rather than passively accepting output, users actively refine, explore, and co-construct text. We conduct a large-scale analysis of this collaborative behavior for users engaged in writing tasks in the wild with two popular AI assistants, Bing Copilot and WildChat. Our analysis goes beyond simple task classification or satisfaction estimation common in prior work and instead characterizes how users interact with LLMs through the course of a session. We identify prototypical behaviors in how users interact with LLMs in prompts following their original request. We refer to these as Prototypical Human-AI Collaboration Behaviors (PATHs) and find that a small group of PATHs explain a majority of the variation seen in user-LLM interaction. These PATHs span users revising intents, exploring texts, posing questions, adjusting style or injecting new content. Next, we find statistically significant correlations between specific writing intents and PATHs, revealing how users' intents shape their collaboration behaviors. We conclude by discussing the implications of our findings on LLM alignment.

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*

[**Link to Conference Handbook**](https://drive.google.com/file/d/1johU5QqVVYO4RfH7QcIORr7qrVBdzdwC/view?usp=sharing)





<br>

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

Keyword decision in Sponsored Search Advertising is critical to the success of ad campaigns. While LLM-based methods offer automated keyword generation, they face three major limitations: reliance on large-scale query–keyword pair data, lack of online multi-objective performance monitoring and optimization, and weak quality control in keyword selection. These issues hinder the agentic use of LLMs in fully automating keyword decisions by monitoring and reasoning over key performance indicators such as impressions, clicks, conversions, and CTA effectiveness. To overcome these challenges, we propose OMS, a keyword generation framework that is On-the-fly (requires no training data, monitors online performance, and adapts accordingly), Multi-objective (employs agentic reasoning to optimize keywords based on multiple performance metrics) and Self-reflective (agentically evaluates keyword quality). Experiments on benchmarks and real-world ad campaigns show that OMS outperforms existing methods; Ablation and human evaluations confirm the effectiveness of each component and the quality of generated keywords.

OMS: On-the-fly, Multi-Objective, Self-Reflective Ad Keyword Generation via LLM Agent

We present the first comprehensive study of Memorization in Multilingual Large Language Models (MLLMs), analyzing 95 languages using models across diverse model scales, architectures, and memorization definitions. As MLLMs are increasingly deployed, understanding their memorization behavior has become critical. Yet prior work has focused primarily on monolingual models, leaving multilingual memorization underexplored, despite the inherently long-tailed nature of training corpora. We find that the prevailing assumption, that memorization is highly correlated with training data availability, fails to fully explain memorization patterns in MLLMs. We hypothesize that treating languages in isolation — ignoring their similarities — obscures the true patterns of memorization. To address this, we propose a novel graph-based correlation metric that incorporates language similarity to analyze cross-lingual memorization. Our analysis reveals that among similar languages, those with fewer training tokens tend to exhibit higher memorization, a trend that only emerges when cross-lingual relationships are explicitly modeled. These findings underscore the importance of a language-aware perspective in evaluating and mitigating memorization vulnerabilities in MLLMs. This also constitutes empirical evidence that language similarity both explains Memorization in MLLMs and underpins Cross-lingual Transferability, with broad implications for multilingual NLP.

Shared Path: Unraveling Memorization in Multilingual LLMs through Language Similarities

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.

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

We introduce Drivelology, a unique linguistic phenomenon characterised as ``nonsense with depth'' -- utterances that are syntactically coherent yet pragmatically paradoxical, emotionally loaded, or rhetorically subversive. While such expressions may resemble surface-level nonsense, they encode implicit meaning requiring contextual inference, moral reasoning, or emotional interpretation. We find that current large language models (LLMs), despite excelling at many NLP tasks, consistently fail to grasp the layered semantics of Drivelological text. To investigate this, we construct a small but diverse benchmark dataset of over 1200 meticulously curated examples with select instances in Mandarin, English, Spanish, French, Japanese, and Korean. Annotation was especially challenging: each of the examples required careful expert review to verify that it truly reflected Drivelological characteristics. The process involved multiple rounds of discussion and adjudication to address disagreements, highlighting the subtle and subjective nature of the Drivelology. We evaluate a range of LLMs on classification, generation, and reasoning tasks. Our results reveal clear limitations: models often confuse Drivelology with shallow nonsense, produce incoherent justifications, or miss the implied rhetorical function altogether. These findings highlight a deeper representational gap in LLMs' pragmatic understanding and challenge the assumption that statistical fluency implies cognitive comprehension. We release our dataset and code to facilitate further research in modelling linguistic depth beyond surface-level coherence.

Drivel-ology: Challenging LLMs with Interpreting Nonsense with Depth

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

Mobile GUI agents have attracted tremendous research participation recently. Traditional approaches to mobile agent training rely on centralized data collection, leading to high cost and limited scalability. Distributed training utilizing federated learning offers an alternative by harnessing real-world user data, providing scalability and reducing costs. However, pivotal challenges, including the absence of standardized benchmarks, hinder progress in this field. To tackle the challenges, we introduce FedMABench, the first benchmark for federated training and evaluation of mobile GUI agents, specifically designed for heterogeneous scenarios. FedMABench features 6 datasets with 30+ subsets, 8 federated algorithms, 10+ base models, and over 800 apps across 5 categories, providing a comprehensive framework for evaluating mobile agents across diverse environments. Through extensive experiments, we uncover several key insights: federated algorithms consistently outperform local training; the distribution of specific apps plays a crucial role in heterogeneity; and, even apps from distinct categories can exhibit correlations during training. FedMABench is publicly available at: https://anonymous.4open.science/r/FedMABench.

FedMABench: Benchmarking Mobile GUI Agents on Decentralized Heterogeneous User Data

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

Language models have demonstrated remarkable capabilities in reasoning tasks through test-time scaling techniques like best-of-N sampling and tree search. However, these approaches often demand substantial computational resources, creating a critical trade-off between performance and efficiency. We introduce STAND (STochastic Adaptive N-gram Drafting), a novel model-free speculative decoding approach that leverages the inherent redundancy in reasoning trajectories to achieve significant acceleration without compromising accuracy. Our analysis reveals that reasoning paths frequently reuse similar reasoning patterns, enabling efficient model-free token prediction without requiring separate draft models. By introducing stochastic drafting and preserving probabilistic information through a memory-efficient logit-based N-gram module, combined with optimized Gumbel-Top-K sampling and data-driven tree construction, STAND significantly improves token acceptance rates. Extensive evaluations across multiple models and reasoning tasks (AIME-2024, GPQA-Diamond, and LiveCodeBench) demonstrate that STAND reduces inference latency by 60-65% compared to standard autoregressive decoding while maintaining accuracy. Furthermore, STAND outperforms state-of-the-art speculative decoding methods by 14-28% in throughput and shows strong performance even in single-trajectory scenarios, reducing inference latency by 48-58%. As a model-free approach, STAND can be applied to any existing language model without additional training, being a powerful plug-and-play solution for accelerating language model reasoning.

Accelerated Test-Time Scaling with Model-Free Speculative Sampling

Sentence representation learning (SRL) aims to learn sentence embeddings that conform to the semantic information of sentences. In recent years, fine-tuning methods based on pre-trained models and contrastive learning frameworks have significantly advanced the quality of sentence representations. However, within the semantic space of SRL models, both word embeddings and sentence representations derived from word embeddings exhibit substantial redundant information, which can adversely affect the precision of sentence representations. Existing approaches predominantly optimize training strategies to alleviate the redundancy problem, lacking fine-grained guidance on reducing redundant representations. This paper proposes a novel approach that dynamically identifies and reduces redundant information from a dimensional perspective, training the SRL model to redistribute semantics on different dimensions, and entailing better sentence representations. Extensive experiments across seven semantic text similarity benchmarks demonstrate the effectiveness and generality of the proposed method. A comprehensive analysis of the experimental results is conducted, and the code/data will be released.

Downloads

Next from EMNLP 2025

OMS: On-the-fly, Multi-Objective, Self-Reflective Ad Keyword Generation via LLM Agent

Stay up to date with the latest Underline news!

PRESENTATIONS

CONFERENCES

COMPANY

RESOURCES