United States

This study investigates the internal representations of verb-particle combinations, called multi-word verbs, within transformer-based large language models (LLMs), specifically examining how these models capture lexical and syntactic properties at different neural network layers. Using the BERT architecture, we analyze the representations of its layers for two different verb-particle constructions: phrasal verbs like &quot;give up&quot; and prepositional verbs like &quot;look at&quot;. Our methodology includes training probing classifiers on the model output to classify these categories at both word and sentence levels. The results indicate that the model’s middle layers achieve the highest classification accuracies. To further analyze the nature of these distinctions, we conduct a data separability test using the Generalized Discrimination Value (GDV). While GDV results show weak linear separability between the two verb types, probing classifiers still achieve high accuracy, suggesting that representations of these linguistic categories may be &quot;non-linearly separable&quot;. This aligns with previous research indicating that linguistic distinctions in neural networks are not always encoded in a linearly separable manner. These findings computationally support usage-based claims on the representation of verb-particle constructions and highlight the complex interaction between neural network architectures and linguistic structures.

NAACL 2025

Probing Internal Representations of Multi-Word Verbs in Large Language Models

data separability

multi-word verbs

probing classifiers

This study investigates the internal representations of verb-particle combinations, called multi-word verbs, within transformer-based large language models (LLMs), specifically examining how these models capture lexical and syntactic properties at different neural network layers. Using the BERT architecture, we analyze the representations of its layers for two different verb-particle constructions: phrasal verbs like "give up" and prepositional verbs like "look at". Our methodology includes training probing classifiers on the model output to classify these categories at both word and sentence levels. The results indicate that the model’s middle layers achieve the highest classification accuracies. To further analyze the nature of these distinctions, we conduct a data separability test using the Generalized Discrimination Value (GDV). While GDV results show weak linear separability between the two verb types, probing classifiers still achieve high accuracy, suggesting that representations of these linguistic categories may be "non-linearly separable". This aligns with previous research indicating that linguistic distinctions in neural networks are not always encoded in a linearly separable manner. These findings computationally support usage-based claims on the representation of verb-particle constructions and highlight the complex interaction between neural network architectures and linguistic structures.

workshop paper

### Welcome to 2025 Annual Conference of the Nations of the Americas Chapter of the Association for Computational Linguistics

Welcome to the 2025 meeting of the Nations of the Americas Chapter of the Association for Computational Linguistics! I am proud to help organize the first NAACL conference to carry the new name of our organization, one that emphasizes inclusion for all of the Americas. I am also pleased to welcome you to Albuquerque, New Mexico, a state whose unique blend of cultural influences will make for an excellent backdrop for NAACL 2025, especially with this year’s special theme on NLP in a Multicultural World. 
**[Continue reading...](https://drive.google.com/file/d/1jX-qGhqVSZZCIrAnJaz798pflu5Irrdn/view?usp=sharing)**

*- Colin Cherry, Google, NAACL 2025 General Chair* 

[![](https://assets.underline.io/markdown_image/1/image/b087f8a4dc5816d6e1a6514e59c59ac3.png)](https://drive.google.com/file/d/1T96GzPqObXrMln2BMByCSXTSizjTg69P/view?usp=sharing)

You need to log in with the email address you registered with. Access credentials have been sent to your email. 

Please be sure to check your spam and other email folders if you do not see an email confirmation right away.

Please log in to explore this event.

To access NAACL 2025 event page you are required to register. Please follow [**this link**](https://2025.naacl.org/registration/registration/) to register. Access will depend on your registration type.

Please register!

Welcome to 2025 Annual Conference of the Nations of the Americas Chapter of the Association for Computational Linguistics

While recent large language models (LLMs) demonstrate remarkable abilities in responding to queries in diverse languages, their ability to handle long multilingual contexts is unexplored. As such, a systematic evaluation of the long-context capabilities of LLMs in multilingual settings is crucial, specifically in the context of information retrieval. To address this gap, we introduce the MultiLingual Needle-in-a-Haystack (MLNeedle) test, designed to assess a model's ability to retrieve relevant information (the needle) from a collection of multilingual distractor texts (the haystack). This test serves as an extension of the multilingual question-answering task, encompassing both monolingual and cross-lingual retrieval. We evaluate four state-of-the-art LLMs on MLNeedle. Our findings reveal that model performance can vary significantly with language and needle position. Specifically, we observe that model performance is the lowest when the needle is (i) in a language outside the English language family, and (ii) located in the middle of the input context. Furthermore, although some models claim a context size of 8k tokens or greater, none demonstrate satisfactory cross-lingual retrieval performance as the context length increases. Our analysis provides key insights into the long-context behavior of LLMs in multilingual settings to guide future evaluation protocols. To our knowledge, this is the first study to investigate the multilingual long-context behavior of LLMs.

Multilingual Needle in a Haystack: Investigating Long-Context Behavior of Multilingual Large Language Models

Generative replay has proven effective in addressing the catastrophic forgetting issue of continual learning (CL) in natural language processing (NLP). However, relying on a single task-specific token or prompt often falls short in generating pseudo-samples that accurately reflect the true data distribution. This leads to issues of semantic inconsistency and scale inconsistency. To tackle these challenges, we propose a Prototype Conditioned Generative Replay (PCGR) method, which enhances generative reply by incorporating task-level statistics through a Prototype Conditioned Variational Autoencoder (PCVAE). Specifically, task-level embedding statistics are stored as prototypes for each old task. When a new task is introduced, PCVAE draws samples from task-specific prototype-based distributions to generate pseudo-samples. By incorporating the prototype, the generated pseudo-samples are both more representative and sufficiently diverse to reflect the real data distribution. Furthermore, as previously stored prototypes may become outdated due to evolving model parameters, we propose a Prototype Shift Estimation (PSE) to adjust for these changes. Experiments on NLP tasks across two different scenarios show that PCGR outperforms previous state-of-the-art (SOTA) methods.

Prototype Conditioned Generative Replay for Continual Learning in NLP

Reasoning and linguistic skills form the cornerstone of human intelligence, facilitating problem-solving and decision-making. Recent advances in Large Language Models (LLMs) have led to impressive linguistic capabilities and emergent reasoning behaviors, fueling widespread adoption across application domains. However, LLMs still struggle with complex reasoning tasks, highlighting their systemic limitations. In this work, we focus on evaluating whether LLMs have the requisite representations to reason using two foundational relationships: ''equivalence'' and ''inheritance''. We introduce novel tasks and benchmarks spanning six languages and observe that current SOTA LLMs often produce conflicting answers to the same questions across languages in 17.3-57.5% of cases and violate inheritance constraints in up to 37.2% cases. To enhance consistency across languages, we propose novel ''Compositional Representations'' where tokens are represented as composition of equivalent tokens across languages, with resulting conflict reduction (up to -4.7%) indicating benefits of shared LLM representations.

Towards Robust Knowledge Representations in Multilingual LLMs for Equivalence and Inheritance based Consistent Reasoning

Large language models (LLMs) have shown impressive capabilities across diverse settings, but still struggle as the length and complexity of the context increases. To address this challenge, we propose Thinking Recursively and Dynamically (ThReaD). THREAD frames model generation as a thread of execution that, based on the context, can run to completion or dynamically spawn new threads. By spawning, threads can offload work (e.g., thinking, retrieving information) to child threads, which only return tokens needed for the parent thread to do its work. We apply THREAD in the settings of LLM task solving and question answering, where the dynamic threading allows the model to recursively decompose the given task or question into progressively simpler sub-problems that can be solved by separate child threads. We test THREAD, implemented using a few-shot learning approach, on diverse benchmarks for agent tasks and data-grounded question answering. THREAD achieves state-of-the-art performance with GPT-4 and GPT-3.5 on these benchmarks, including ALFWorld, TextCraft, and WebShop, along with two new benchmarks, DataCommons QA and MIMIC-III ICU QA. In addition, THREAD outperforms existing frameworks by 10% to 50% absolute points with smaller models, including Llama-3-8b and CodeLlama-7b.

THREAD: Thinking Deeper with Recursive Spawning

The prevalence of toxic behavior in online gaming communities necessitates robust detection methods to ensure user safety. We introduce GameTox, a novel dataset comprising 53K game chat utterances annotated for toxicity detection through intent classification and slot filling. This dataset captures the complex relationship between user intent and specific linguistic features that contribute to toxic interactions. We extensively analyze the dataset to uncover key insights into the nature of toxic speech in gaming environments. Furthermore, we establish baseline performance metrics using state-of-the-art natural language processing and large language models, demonstrating the dataset's contribution towards enhancing the detection of toxic behavior and revealing the limitations of contemporary models. Our results indicate that leveraging both intent detection and slot filling provides a significantly more granular and context-aware understanding of harmful messages. This dataset serves as a valuable resource to train advanced models that can effectively mitigate toxicity in online gaming and foster healthier digital spaces. Our dataset is publicly available at: https://github.com/shucoll/GameTox.

GameTox: A Comprehensive Dataset and Analysis for Enhanced Toxicity Detection in Online Gaming Communities

Data is the lifeblood of large language models (LLMs). While the quantity of open-source data available for training LLMs is substantial, its integrity often falls short. For instance, the open-source chat version of Yi-1.5-9B scores 5.20 on AlignBench, while the Chinese Alpaca-GPT4 version scores 4.12. This discrepancy makes it challenging for developers to create models that excel in downstream tasks and instruction following. Therefore, it is essential to improve data integrity. Currently, there are two mainstream methods for enhancing data integrity: data filtering and data augmentation. Due to the labor-intensive and time-consuming nature of performing these tasks manually, some of these efforts are now being undertaken by LLMs, owing to their high alignment with human preferences. However, we have found that performing data filtering or data augmentation with LLMs has limited effectiveness in improving data integrity. In this work, we propose FiNE (**F**iltering and **I**mproving **N**oisy data **E**laborately), a method that performs refined filtering and improvement of training data with LLMs. Using the data obtained through our method to train Yi-1.5-9B, the performance gap on AlignBench between our model and the open-source chat version is reduced from 1.08 to 0.35. Additionally, on HalluQA, our model surpasses the open-source chat version by 8.45.

FiNE: Filtering and Improving Noisy Data Elaborately with Large Language Models

The effectiveness of embedding methods is crucial for optimizing text classification performance in Automated Machine Learning (AutoML). However, selecting the most suitable pre-trained model for a given task remains challenging. This study introduces the Corpus-Driven Domain Mapping (CDDM) pipeline, which utilizes a domain-annotated corpus of pre-fine-tuned models from the Hugging Face Model Hub to improve model selection. Integrating these models into AutoML systems significantly boosts classification performance across multiple datasets compared to baseline methods. Despite some domain recognition inaccuracies, results demonstrate CDDM’s potential to enhance model selection, streamline AutoML workflows, and reduce computational costs.

AutoML Meets Hugging Face: Domain-Aware Pretrained Model Selection for Text Classification

This paper presents the construction of VIDiom-PT, a corpus in European Portuguese annotated for verbal idioms (e.g. O Rui bateu a bota, lit.: Rui hit the boot `Rui died'). This linguistic resource aims to support the development of systems capable of processing such constructions in this language variety. To assist in the annotation effort, two tools were built. The first allows for the detection of possible instances of verbal idioms in texts, while the second provides a graphical interface for annotating them. This effort culminated in the annotation of a total of 5,178 instances of 747 different verbal idioms in more than 200,000 sentences in European Portuguese. A highly reliable inter-annotator agreement was achieved, using Krippendorff's alpha for nominal data (0.869) with 5% of the data independently annotated by 3 experts. Part of the annotated corpus is also made publicly available.

A European Portuguese corpus annotated for verbal idioms

Language models are able to handle compositionality and, to some extent, non-compositional phenomena such as semantic idiosyncrasy --a feature most prominent in the case of idioms. This work presents the MultiCoPIE corpus that includes potentially idiomatic expressions as well as non-ambiguous idioms in Catalan, Italian, and Russian, extending language coverage of PIE corpus data. The new corpus provides additional information on idiom features, such as their semantic transparency, part-of-speech of idiom head as well as their idiomatic counterparts in English. With this new resource at hand, we first fine-tune an XLM-RoBERTa model to classify figurative and literal usage of potentially idiomatic expressions in English and then study cross-lingual transfer to the languages represented in MultiCoPIE to evaluate the model’s capability to generalize an idiom-related task to other languages not observed during fine-tuning. We show the effect of ‘cross-lingual lexical overlap’: the performance of the model, fine-tuned on English idiomatic expressions and tested on the MultiCoPIE languages, increases significantly when classifying idiomatic expressions that have direct counterparts in English. While this may raise concerns regarding cross-lingual learning and generalization, the outcomes of the experiment on PIEs without English equivalents provide clear evidence of cross-lingual transfer.

MultiCoPIE: A Multilingual Corpus of Potentially Idiomatic Expressions for Cross-lingual PIE Disambiguation

The Irish language has been deemed ‘definitely endangered’ (Moseley, 2012) and has been classified as having ‘weak or no support’ (Lynn, 2023) regarding digital resources in spite of its status as the first official and national language of the Republic of Ireland. This research develops the first named entity recognition (NER) tool for the Irish language, one of the essential tasks identified by the Digital Plan for Irish (Ní Chasaide et al., 2022). In this study, we produce a small gold-standard NER-annotated corpus and compare both monolingual and multilingual BERT models fine-tuned on this task. We experiment with different model architectures and low-resource language approaches to enrich our dataset. We test our models on a mix of single- and multi-word named entities as well as a specific multi-word named entity test set. Our proposed gaBERT model with the implementation of random data augmentation and a conditional random fields layer demonstrates significant performance improvements over baseline models, alternative architectures, and multilingual models, achieving an F1 score of 76.52. This study contributes to advancing Irish language technologies and supporting Irish language digital resources, providing a basis for Irish NER and identification of other MWE types.

Content not yet available

Downloads

Next from NAACL 2025

Multilingual Needle in a Haystack: Investigating Long-Context Behavior of Multilingual Large Language Models

Stay up to date with the latest Underline news!

PRESENTATIONS

CONFERENCES

COMPANY

RESOURCES

Content not yet available

.css-70qvj9{display:-webkit-box;display:-webkit-flex;display:-ms-flexbox;display:flex;-webkit-align-items:center;-webkit-box-align:center;-ms-flex-align:center;align-items:center;}Downloads

Next from NAACL 2025

Multilingual Needle in a Haystack: Investigating Long-Context Behavior of Multilingual Large Language Models

Stay up to date with the latest Underline news!

PRESENTATIONS

CONFERENCES

COMPANY

RESOURCES

Downloads