Austria

Despite the fact that large language models (LLMs) show exceptional skill in instruction following tasks, this strength can turn into a vulnerability when the models are required to disregard certain instructions. Instruction following tasks typically involve a clear task description and input text containing the target data to be processed. However, when the input itself resembles an instruction, confusion may arise, even if there is explicit prompting to distinguish between the task instruction and the input. We refer to this phenomenon as instructional distraction. In this paper, we introduce a novel benchmark, named **DIM-Bench**, specifically designed to assess LLMs&#39; performance under instructional distraction. The benchmark categorizes real-world instances of instructional distraction and evaluates LLMs across four instruction tasks: proofreading, rewriting, translation, and style transfer---alongside five input tasks: reasoning, code generation, mathematical reasoning, bias detection, and question answering. Our experimental results reveal that even the most advanced LLMs are susceptible to instructional distraction, often failing to accurately follow user intent in such cases.

ACL 2025

LLMs can be easily Confused by Instructional Distractions

instruction-following

llms

benchmark

evaluation

### 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!

NAACL 2025

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

Despite the fact that large language models (LLMs) show exceptional skill in instruction following tasks, this strength can turn into a vulnerability when the models are required to disregard certain instructions. Instruction following tasks typically involve a clear task description and input text containing the target data to be processed. However, when the input itself resembles an instruction, confusion may arise, even if there is explicit prompting to distinguish between the task instruction and the input. We refer to this phenomenon as instructional distraction. In this paper, we introduce a novel benchmark, named **DIM-Bench**, specifically designed to assess LLMs' performance under instructional distraction. The benchmark categorizes real-world instances of instructional distraction and evaluates LLMs across four instruction tasks: proofreading, rewriting, translation, and style transfer---alongside five input tasks: reasoning, code generation, mathematical reasoning, bias detection, and question answering. Our experimental results reveal that even the most advanced LLMs are susceptible to instructional distraction, often failing to accurately follow user intent in such cases.

poster

### Welcome to The 63rd Annual Meeting of the Association for Computational Linguistics (ACL 2025)

Message from the General Chair: 
*It is my great pleasure and honor to welcome you to the 63rd Annual Meeting of the Association for Computational Linguistics (ACL 2025), held in beautiful Vienna, Austria, from July 27 to August 1, 2025. ACL2025continues our field’s tradition of excellence in scholarship, innovation, and inclusivity, and I am deeply grateful to the many volunteers who have worked tirelessly to bring this event to life.* 
[Read more](https://drive.google.com/file/d/1GI_hvOpjswAuYdUTromfeDiPpCcqidwg/view?usp=sharing)

To access this event page, you need to log in with the **email address you registered with**. Access credentials will be sent to your email from Underline - subject line "Welcome to ACL 2025". 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 need to log in with the **email address you registered with**. 

Welcome to The 63rd Annual Meeting of the Association for Computational Linguistics (ACL 2025)

The rapid advancement of Large Language Models (LLMs) poses significant challenges for safety evaluation. Current static datasets struggle to identify emerging vulnerabilities due to three limitations: (1) they risk being exposed in model training data, leading to evaluation bias; (2) their limited prompt diversity fails to capture real-world application scenarios; (3) they are limited to provide human-like multi-turn interactions. To address these limitations, we propose a dynamic evaluation framework, CogSafe, for comprehensive and automated multi-turn safety assessment of LLMs. We introduce CogSafe based on cognitive theories to simulate the real chatting process. To enhance assessment diversity, we introduce scenario simulation and strategy decision to guide the dynamic generation, enabling coverage of application situations. Furthermore, we incorporate the cognitive process to simulate multi-turn dialogues that reflect the cognitive dynamics of real-world interactions. Extensive experiments demonstrate the scalability and effectiveness of our framework, which has been applied to evaluate the safety of widely used LLMs.

Dynamic Evaluation with Cognitive Reasoning for Multi-turn Safety of Large Language Models

In modern large language models (LLMs), LLM alignment is of crucial importance and is typically achieved through methods such as reinforcement learning from human feedback (RLHF) and direct preference optimization (DPO). However, in most existing methods for LLM alignment, all tokens in the response are optimized using a sparse, response-level reward or preference annotation. The ignorance of token-level rewards may erroneously punish high-quality tokens or encourage low-quality tokens, resulting in suboptimal performance and slow convergence speed. To address this issue, we propose \emph{\textbf{AlignDistil}}, a RLHF-equivalent distillation method for token-level reward optimization. Specifically, we introduce the reward learned by DPO into the RLHF objective and theoretically prove the equivalence between this objective and a token-level distillation process, where the teacher distribution linearly combines the logits from the DPO model and a reference model. On this basis, we further bridge the accuracy gap between the reward from the DPO model and the pure reward model, by building a contrastive DPO reward with a normal and a reverse DPO model. Moreover, to avoid under- and over-optimization on different tokens, we design a token adaptive logit extrapolation mechanism to construct an appropriate teacher distribution for each token. Experimental results demonstrate the superiority of our AlignDistil over existing methods and showcase fast convergence due to its token-level distributional reward optimization.

AlignDistil: Token-Level Language Model Alignment as Adaptive Policy Distillation

In language model training, it is desirable to equip models with capabilities from various tasks. However, it is not clear how to directly obtain the right data mixtures for these capabilities as the relationship between data and tasks is difficult to be modeled. In this work, we observe that checkpoint models exhibit emerging capabilities at different points in the training trajectory. Often, the training process saves checkpoints as artifacts that are under-utilized as a source of in-training data signals. We identify these artifact models based on their respective capabilities on the benchmarks and leverage them as data mixers by using their aggregated first-order influence approximation over source data. We demonstrated on eight reasoning benchmarks that the proposed framework shows significant improvements in the pretraining setting, with accuracy increases of up to 1.93%. Overall, this demonstrates the potential of checkpoint models to enhance data quality and optimize data mixtures.

AutoMixer: Checkpoint Artifacts as Automatic Data Mixers

In this work, we tackle the challenge of embedding realistic human personality traits into LLMs. Previous approaches have primarily focused on prompt-based methods that describe the behavior associated with the desired personality traits, suffering from realism and validity issues. To address these limitations, we introduce BIG5-CHAT, a large-scale dataset containing 100,000 dialogues designed to ground models in how humans express their personality in text. Leveraging this dataset, we explore Supervised Fine-Tuning and Direct Preference Optimization as training-based methods to align LLMs more naturally with human personality patterns. Our methods outperform prompting on personality assessments such as BFI and IPIP-NEO, with trait correlations more closely matching human data. Furthermore, our experiments reveal that models trained to exhibit higher conscientiousness, higher agreeableness, lower extraversion, and lower neuroticism display better performance on reasoning tasks, aligning with psychological findings on how these traits impact human cognitive performance. To our knowledge, this work is the first comprehensive study to demonstrate how training-based methods can shape LLM personalities through learning from real human behaviors.

BIG5-CHAT: Shaping LLM Personalities Through Training on Human-Grounded Data

Inference-time intervention (ITI) has emerged as a promising method for steering large language model (LLM) behavior in a particular direction (e.g., improving helpfulness) by intervening on token representations without costly updates to the LLM's parameters. However, existing ITI approaches fail to scale to multi-attribute settings with conflicts, such as enhancing helpfulness while also reducing toxicity. To address this, we introduce Multi-Attribute Targeted Steering (MAT-Steer), a novel steering framework designed for selective token-level intervention across multiple attributes. We achieve this by learning steering vectors using an alignment objective that shifts the model's internal representations of undesirable outputs closer to those of desirable ones while enforcing sparsity and orthogonality among vectors for different attributes, thereby reducing inter-attribute conflicts. We evaluate MAT-Steer in two distinct settings: (i) on question answering (QA) tasks where we balance attributes like truthfulness, bias, and toxicity; (ii) on generative tasks where we simultaneously improve attributes like helpfulness, correctness, and coherence. MAT-Steer outperforms existing ITI and parameter-efficient fine-tuning approaches across both task types (e.g., average 3% accuracy gain across QA tasks and 55.82% win rate against the best ITI baseline).

Multi-Attribute Steering of Language Models via Targeted Intervention

Peer review is fundamental to scientific research, but the growing volume of publications has intensified the challenges of this expertise-intensive process. While LLMs show promise in various scientific tasks, their potential to assist with peer review, particularly in identifying paper limitations, remains understudied. We first present a comprehensive taxonomy of limitation types in scientific research, with a focus on AI. Guided by this taxonomy, for studying limitations, we present LimitGen, the first comprehensive benchmark for evaluating LLMs' capability to support early-stage feedback and complement human peer review. Our benchmark consists of two subsets: LimitGen-Syn, a synthetic dataset carefully created through controlled perturbations of high-quality papers, and LimitGen-Human, a collection of real human-written limitations. To improve the ability of LLM systems to identify limitations, we augment them with literature retrieval, which is essential for grounding identifying limitations in prior scientific findings. Our approach enhances the capabilities of LLM systems to generate limitations in research papers, enabling them to provide more concrete and constructive feedback.

Can LLMs Identify Critical Limitations within Scientific Research? A Systematic Evaluation on AI Research Papers

Crosslingual transfer is crucial to contemporary language models’ multilingual capabilities, but how it occurs is not well understood. We
ask what happens to a monolingual language model when it begins to be trained on a second language. Specifically, we train small bilingual models for which we control the amount of data for each language and the order of language exposure. To find evidence of shared multilingual representations, we turn to structural priming, a method used to study grammatical representations in humans. We first replicate previous crosslingual structural priming results and find that after controlling for training data quantity and language exposure, there are asymmetrical effects across language pairs and directions. We argue that this asymmetry may shape hypotheses about human structural priming effects. We also find that structural priming effects are less robust for less similar language pairs, highlighting potential limitations of crosslingual transfer learning and shared representations for typologically diverse languages.

On the Acquisition of Shared Grammatical Representations in Bilingual Language Models

Language is an intricately structured system, and a key goal of NLP interpretability is to provide methodological insights for understanding how language models internally represent this structure. In this paper, we use Shapley Taylor interaction indices (STII) in order to examine how language and speech models internally relate and structure their inputs. Pairwise Shapley interactions give us an attribution measure of how much two inputs work together to influence model outputs beyond if we linearly added their independent influences, providing a view into how models encode structural interactions between inputs. We relate the interaction patterns in models to three underlying linguistic structures: syntactic structure, non-compositional semantics, and phonetic interaction. We find that autoregressive text models encode interactions that correlate with the syntactic proximity of inputs, and that both autoregressive and masked models encode nonlinear interactions in idiomatic phrases with non-compositional semantics. Our speech results show that inputs are more entangled for pairs where a neighboring consonant is likely to influence a vowel or approximant, showing that models encode the phonetic interaction needed for extracting discrete phonemic representations.

Using Shapley interactions to understand how models use structure

Current LLM pipelines account for only one possible tokenization for a given string, ignoring exponentially many alternative tokenizations during training and inference. For example, the $\texttt{Llama3}$ standard tokenization of penguin is $\texttt{[p,enguin]}$, yet $\texttt{[peng,uin]}$ is another perfectly valid alternative. In this paper, we show that despite LLMs being trained solely on one tokenization, they still retain semantic understanding of other tokenizations, raising questions about their implications in LLM safety. Put succinctly, we answer the following question: can we adversarially tokenize an obviously malicious string to evade safety and alignment restrictions? We show that not only is adversarial tokenization an effective yet previously neglected axis of attack, but it is also competitive against existing state-of-the-art adversarial approaches without changing the text of the harmful request. We empirically validate this exploit across three state-of-the-art LLMs and adversarial datasets, revealing a previously unknown vulnerability in subword models.

Adversarial Tokenization

Figurative language (e.g., irony, hyperbole, understatement) is ubiquitous in human communication, resulting in utterances where the literal and the intended meanings do not match. The Rational Speech Act (RSA) framework, which explicitly models speaker intentions, is the most widespread theory of probabilistic pragmatics, but existing implementations are either unable to account for figurative expressions or require modeling the implicit motivations for using figurative language (e.g., to express joy or annoyance) in a setting-specific way. In this paper, we introduce the Rhetorical-Strategy-Aware RSA (RSA)^2 framework which models figurative language use by considering a speaker’s employed rhetorical strategy. We show that (RSA)^2 enables human-compatible interpretations of non-literal utterances without modeling a speaker’s motivations for being non-literal. Combined with LLMs, it achieves state-of-the-art performance on the ironic split of PragMega+, a new irony interpretation dataset introduced in this study.

Downloads

Next from ACL 2025

Dynamic Evaluation with Cognitive Reasoning for Multi-turn Safety of Large Language Models

Similar lecture

Find the Intention of Instruction: Comprehensive Evaluation of Instruction Understanding for Large Language Models