India

Tokenization is the first step in training any Large Language Model (LLM), where the text is split into a sequence of tokens as per the model&#39;s fixed vocabulary. This tokenization in LLMs is different from the traditional tokenization in NLP where the text is split into a sequence of &quot;natural&quot; words. In LLMs, a natural word may also be broken into multiple tokens due to limited vocabulary size of the LLMs (e.g., Mistral&#39;s tokenizer splits &quot;martial&quot; into &quot;mart&quot; and &quot;ial&quot;). In this paper, we hypothesize that such breaking of natural words negatively impacts LLM performance on various NLP tasks. To quantify this effect, we propose a set of penalty functions that compute a tokenization penalty for a given text for a specific LLM, indicating how &quot;bad&quot; the tokenization is. We establish statistical significance of our hypothesis on multiple NLP tasks for a set of different LLMs.

IJCNLP-AACL 2025

Broken Words, Broken Performance: Effect of Tokenization on Performance of LLMs

nlp tasks

tokenization

large language models

## Welcome to EMNLP 2024! 
We are excited to welcome you to one of the most prominent conferences in the field of Natural Language Processing. This year, EMNLP 2024 is being held in a hybrid format,
offering both virtual and in-person participation in beautiful Miami. Due to a record-breaking number of submissions, we've expanded the total number of accepted papers to accommodate more cutting-edge research from around the globe.
### [Conference Handbook](https://drive.google.com/file/d/1WPROgxjLAC96AJL7Ugy0tEnYm7dkrbHt/view?usp=sharing)

You are required to register for this event. **Please register [here](https://2024.emnlp.org/registration/).** The EMNLP 2024 event page on Underline will be open to public one week prior to the event.

Please register!

EMNLP 2024

EMNLP 2024 will take place in Miami, Florida from Nov 12th to Nov 16th, 2024, at the Hyatt Regency Miami Hote and on Underline for remote participants.

Tokenization is the first step in training any Large Language Model (LLM), where the text is split into a sequence of tokens as per the model's fixed vocabulary. This tokenization in LLMs is different from the traditional tokenization in NLP where the text is split into a sequence of "natural" words. In LLMs, a natural word may also be broken into multiple tokens due to limited vocabulary size of the LLMs (e.g., Mistral's tokenizer splits "martial" into "mart" and "ial"). In this paper, we hypothesize that such breaking of natural words negatively impacts LLM performance on various NLP tasks. To quantify this effect, we propose a set of penalty functions that compute a tokenization penalty for a given text for a specific LLM, indicating how "bad" the tokenization is. We establish statistical significance of our hypothesis on multiple NLP tasks for a set of different LLMs.

technical paper

### Welcome to IJCNLP-AACL 2025! 
 It is a great honor to host this joint conference in Mumbai, India, from December 20 to 24, 2025. The joint conferences of IJCNLP and AACL are organized with alternating leadership in the Asia-Pacific region. The event is run by the Asian Federation of Natural Language Processing (AFNLP) in odd years, and by AACL in even years, while it is organized solely by ACL when the annual ACL meeting is held in the region. This year, the conference is primarily organized by AFNLP. 
*Kentaro Inui
MBZUAI, UAE
General Chair, IJCNLP-AACL 2025* 
Read full message and download the Conference Handbook [**here**](https://drive.google.com/file/d/1UTwxkAqSqI-GAoJC3wE1zZt5VP1Y8GX0/view?usp=sharing).

The 14th IJCNLP & 4th AACL will be held in Mumbai, India from December 20th to December 24th, 2025.

Using Large Language Models (LLMs) to simulate user opinions has received growing attention. Yet LLMs, especially trained with reinforcement learning from human feedback (RLHF), are known to exhibit biases toward dominant viewpoints, raising concerns about their ability to represent users from diverse demographic and cultural backgrounds. In this work, we examine the extent to which LLMs can simulate human responses to cross-domain survey questions and propose two LLM-based approaches: chain-of-thought (COT) prompting and Diverse Claims Generation (CLAIMSIM), which elicits viewpoints from LLM parametric knowledge as contextual input. Experiments on the survey question answering task indicate that, while CLAIMSIM produces more diverse responses, both approaches struggle to accurately simulate users. Further analysis reveals two key limitations: (1) LLMs tend to maintain fixed viewpoints across varying demographic features, and generate single-perspective claims; and (2) when presented with conflicting claims, LLMs struggle to reason over nuanced differences among demographic features, limiting their ability to adapt responses to specific user profiles.

An Analysis of Large Language Models for Simulating User Responses in Surveys

Quotation recommendation is an inherently asymmetric retrieval task, where the intended meaning of a quote often diverges from surface expressions, creating significant semantic shifts. Combined with minimal lexical overlap, this poses a core challenge for classic dense retrievers, which struggle to capture non-literal and rhetorical alignments. To bridge this semantic gap, we propose introducing controllable signals to guide the model’s attention toward abstract, context-relevant concepts. We propose CtrlShift, a framework that leverages a Variational Autoencoder (VAE) to capture latent associations between context and quotation, which is used to derive context-aware control signals to modulate semantic focus and support bidirectional alignment and rhetorical intent modeling. Experiments show that our method consistently outperforms baselines on the quotation recommendation task and can be effectively transfered to the general purposed benchmark. Further, CtrlShift integrates seamlessly with general-purpose generative models without additional fine-tuning, and provides satisfactory interpretability by generating textual explaination to uncover the model’s focus on abstract, citation-aligned semantics.

CtrlShift: Steering Language Models for Dense Quotation Retrieval with Dynamic Prompts

Large Language Models (LLMs) demonstrate strong multilingual capabilities but are costly to deploy due to their size and computational demands. To mitigate this, compression techniques such as pruning and quantization are widely used. However, these methods face two key limitations: (1) they assume access to high-quality instruction or calibration data, which is often unavailable for low-resource languages; and (2) they aim to preserve multilingual generality, making them inefficient for language-specific applications. We introduce LangCompress, a language-aware compression framework that enhances existing compression methods for targeted deployment. LangCompress is method-agnostic and improves state-of-the-art pruning and quantization approaches. It features two core components: an iterative self-supervised pipeline for generating instruction data in the target language, and a vocabulary simplification strategy that reduces the LM head to focus on key tokens. Experiments on perplexity, translation, and summarization tasks show that LangCompress improves performance in the target language. The code and data are publicly available.

LangCompress: Language-Aware Compression of Large Language Models

A stereotype is a generalised claim about a social group. Such claims change with culture and context and are often phrased in everyday language, which makes them hard to detect: the State of the Art Large Language Models (LLMs) reach only 68\% macro-F1 on the yes/no task “does this sentence contain a stereotype?”.  We present HEARTS, a Holistic framework for Explainable, sustAinable and Robust Text Stereotype detection that brings together NLP and social-science. The framework is built on the Expanded Multi-Grain Stereotype Dataset (EMGSD), 57\,201 English sentences that cover gender, profession, nationality, race, religion and LGBTQ+ topics, adding 10\% more data for under-represented groups while keeping high annotator agreement ($\kappa = 0.82$). Fine-tuning the lightweight ALBERT-v2 model on EMGSD raises binary detection scores to 81.5\% macro-F1, matching full BERT while producing 200$\times$ less CO$_2$. For Explainability, we blend SHAP and LIME token level scores and introduce a confidence measure that increases when the model is correct ($\rho = 0.18$). We then use HEARTS to assess 16 SOTA LLMs on 1050 neutral prompts each for stereotype propagation: stereotype rates fall by 23\% between model generations, yet clear differences remain across model families (LLaMA $>$ Gemini $>$ GPT $>$ Claude).  HEARTS thus supplies a practical, low-carbon and interpretable toolkit for measuring stereotype bias in language.

HEARTS: A Holistic Framework for Explainable, Sustainable and Robust Text Stereotype Detection

Security-related data often exists in complex, multi-table formats and is scarce due to privacy and compliance constraints—posing a major challenge for training and evaluating language models (LMs) on security reasoning tasks. In this work, we systematically investigate the performance of large language models (LLMs) across different parameter scales in generating and solving multi-step, semantically rich queries over realistic security scenarios represented through three interlinked tabular datasets. We assess models on three key axes (i) their ability to formulate insightful, high complexity security questions; (ii) the quality and coherence of their reasoning chains; and (iii) their accuracy in deriving actionable answers from the underlying data. To address data scarcity, we propose a diffusion-based synthetic data generation pipeline that amplifies the existing dataset while preserving domain semantics and statistical structure. Our findings reveal that while large models often outperform in reasoning depth and query formulation, smaller models show surprising efficiency and accuracy. The study provides actionable insights for deploying generative models in security analytics and opens avenues for synthetic data-driven evaluation of LLMs in low-resource, high-stakes domains.

Comparing Language Models of Different Scales for Security-Focused Tabular Query Generation and Reasoning

Journalistic interviews are creative, dynamic processes where success hinges on insightful, real-time questioning. While Large Language Models (LLMs) can assist, their tendency to generate coherent but uninspired questions optimizes for probable, not insightful, continuations. This paper investigates whether a structured, multi-agent approach can overcome this limitation to act as a more effective creative partner for journalists. We introduce MAJI, a system designed for this purpose, which employs a divergent-convergent architecture: a committee of specialized agents generates a diverse set of questions, and a convergent agent selects the optimal one. We evaluated MAJI against a suite of strong LLM baselines. Our results demonstrate that our multi-agent framework produces questions that are more coherent, elaborate, and original (+36.9% for our best model vs. a standard LLM baseline), exceeded strong LLM baselines on key measures of creative question quality. Most critically, in a blind survey, professional journalists preferred MAJI's selected questions over those from the baseline by a margin of more than two to one. We present the system's evolution, highlighting the architectural trade-offs that enable MAJI to augment, rather than simply automate, journalistic inquiry. We will release the code upon publication.

MAJI: A Multi-Agent Workflow for Augmenting Journalistic Interviews

Thomas C. Schelling, awarded the 2005 Nobel Memorial Prize in Economic Sciences, pointed out that ''individuals decisions (micromotives), while often personal and localized, can lead to societal outcomes (macrobehavior) that are far more complex and different from what the individuals intended.'' The current research related to large language models' (LLMs') micromotives, such as preferences or biases, assumes that users will make more appropriate decisions once LLMs are devoid of preferences or biases. Consequently, a series of studies has focused on removing bias from LLMs. In the NLP community, while there are many discussions on LLMs' micromotives, previous studies have seldom conducted a systematic examination of how LLMs may influence society's macrobehavior. In this paper, we follow the design of Schelling's model of segregation to observe the relationship between the micromotives and macrobehavior of LLMs. Our results indicate that, regardless of the level of bias in LLMs, a highly segregated society will emerge as more people follow LLMs' suggestions. We hope our discussion will spark further consideration of the fundamental assumption regarding the mitigation of LLMs' micromotives and encourage a reevaluation of how LLMs may influence users and society.

Observing Micromotives and Macrobehavior of Large Language Models

The potential of Large Language Models (LLMs) as instruments for measuring social phenomena is constrained by the methodological limitations of current probing techniques. Prevailing methods rely on static, handcrafted probe sets whose quality is highly dependent on their authors' subjective expertise. This results in measurement tools with inconsistent statistical reliability that defy systematic optimization. Such an "artisanal" approach, akin to using an "uneven ruler," undermines the scientific rigor of its findings and severely limits the applicability of LLMs in the social sciences. To elevate bias measurement from a craft to a science, we introduce the Psychometric-driven Probe Optimization (PMPO) framework. This framework treats a probe set as an optimizable scientific instrument and, for the first time, utilizes a Neural Genetic Algorithm that leverages a powerful LLM as a "neural genetic operator." Through a hybrid strategy of gradient-guided mutation and creative rephrasing, PMPO automatically enhances the probe set's reliability, sensitivity, and diversity. We first establish the external validity of our foundational measurement method (PLC), demonstrating a high correlation between its measurement of occupational gender bias and real-world U.S. Bureau of Labor Statistics data (average Pearson's r=0.83, p<.001). Building on this, we show that the PMPO framework can elevate a standard probe set's internal consistency (Cronbach's Alpha) from 0.90 to an exceptional 0.96 within 10 generations. Critically, in a rigorous, double-blind "Turing Test," probes evolved by PMPO from non-expert seeds were judged by sociology experts to have achieved a level of quality, sophistication, and nuance that is comparable to, and even indistinguishable from, those handcrafted by domain experts. This work provides a systematic pathway to upgrade LLM measurement tools from artisanal artifacts to automated scientific instruments, offering an unprecedented and trustworthy tool for AI safety auditing and computational social science.

PMPO: A Self-Optimizing Framework for Creating High-Fidelity Measurement Tools for Social Bias in Large Language Models

Survey papers play a critical role in scientific communication by consolidating progress across a field. Recent advances in Large Language Models (LLMs) offer a promising solution by automating key steps in the survey-generation pipeline, such as retrieval, structuring, and summarization. However, existing LLM-based approaches often struggle with maintaining coherence across long, multi-section surveys and providing comprehensive citation coverage. To address these limitations,  we introduce SurveyGen-I, an automatic survey generation framework that combines coarse-to-fine retrieval, adaptive planning, and memory-guided generation. SurveyGen-I performs survey-level retrieval to construct the initial outline and writing plan, then dynamically refines both during generation through a memory mechanism that stores previously written content and terminology, ensuring coherence across subsections. When the system detects insufficient context, it triggers fine-grained subsection-level retrieval. During generation, SurveyGen-I leverages this memory mechanism to maintain coherence across subsections. Experiments across six scientific domains demonstrate that SurveyGen-I consistently outperforms previous works in content quality, consistency, and citation coverage. The code is available at https://github.com/SurveyGens/SurveyGen-I.

SurveyGen-I: Consistent Scientific Survey Generation with Evolving Plans and Memory-Guided Writing

While state-of-the-art large language models find high rates of success on text classification tasks such as sentiment analysis, they still exhibit vulnerabilities to adversarial examples: meticulously crafted perturbations of input data that guide models into making false predictions. These adversarial attacks are of particular concern when the systems in question are user-facing. While many attacks are able to reduce the accuracy of Transformer-based classifiers by a substantial margin, they often require a large amount of computational time and a large number of queries made to the attacked model, which makes the attacks susceptible to detection. In this work, we resolve the limitations of high query counts and necessary computational time by proposing a query-efficient word-level attack that is fast during deployment and does not compromise the attack success rate of state-of-the-art methods. Our attack constructs a dictionary of adversarial word substitutions based on prior data and leverages these substitutions to flip the sentiment classification of the text. Our attack method achieves an average of 27.49 queries—over 30% fewer than the closest competitor—while maintaining a 99.70% attack success rate. We also develop an effective defense strategy inspired by our attack approach.

Downloads

Next from IJCNLP-AACL 2025

An Analysis of Large Language Models for Simulating User Responses in Surveys

Similar lecture

Tokenization Is More Than Compression