Austria

Large language models (LLMs) are known to have the potential to generate harmful content, posing risks to users. While significant progress has been made in developing taxonomies for LLM risks and safety evaluation prompts, most studies have focused on monolingual contexts, primarily in English. However, language- and region-specific risks in bilingual contexts are often overlooked, and core findings can diverge from those in monolingual settings. In this paper, we introduce Qorǵau, a novel dataset specifically designed for safety evaluation in Kazakh and Russian, reflecting the unique bilingual context in Kazakhstan, where both Kazakh (a low-resource language) and Russian (a high-resource language) are spoken. Experiments with both multilingual and language-specific LLMs reveal notable differences in safety performance, emphasizing the need for tailored, region-specific datasets to ensure the responsible and safe deployment of LLMs in countries like Kazakhstan. Warning: this paper contains example data that may be offensive, harmful, or biased.

ACL 2025

Qorǵau: Evaluating Safety in Kazakh-Russian Bilingual Contexts

multilingual corpora

automatic evaluation of datasets

language resources

evaluation

### 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)

poster

In this paper, we introduce the Akan Cinematic Emotions (AkaCE) dataset, the first multimodal emotion dialogue dataset for an African language, addressing the significant lack of resources for low-resource languages in emotion recognition research. AkaCE, developed for the Akan language, contains 385 emotion-labeled dialogues and 6162 utterances across audio, visual, and textual modalities, along with word-level prosodic prominence annotations. The presence of prosodic labels in this dataset also makes it the first prosodically annotated African language dataset. We demonstrate the quality and utility of AkaCE through experiments using state-of-the-art emotion recognition methods, establishing solid baselines for future research. We hope AkaCE inspires further work on inclusive, linguistically and culturally diverse NLP resources.

Akan Cinematic Emotions (ACE): A Multimodal Multi-party Dataset for Emotion Recognition in Movie Dialogues

Large Language Models (LLMs) can transfer their reasoning skills to smaller models by teaching them to generate the intermediate reasoning process required to solve multistep reasoning tasks. While LLMs can accurately solve reasoning tasks through a variety of strategies, even without fine-tuning, smaller models are not expressive enough to fit the LLMs distribution on all strategies when distilled and tend to prioritize one strategy over the others. This reliance on one strategy poses a challenge for smaller models when attempting to solve reasoning tasks that may be difficult with their preferred strategy. To address this, we propose a distillation method SIKeD: **S**elf-guided **I**terative **K**nowledg**e** **D**istillation, where the LLM teaches the smaller model to approach a task using different strategies and the smaller model uses its self-generated on-policy outputs to choose the most suitable strategy for the given task. The training continues in a self-guided iterative manner, where for each training iteration, a decision is made on how to combine the LLM data with the self-generated outputs. Unlike traditional distillation methods, SIKeD allows the smaller model to learn which strategy is suitable for a given task while continuously learning to solve a task using different strategies. Our experiments on various mathematical reasoning datasets show that SIKeD significantly outperforms traditional distillation techniques across smaller models of different sizes.

SIKeD: Self-guided Iterative Knowledge Distillation for Mathematical Reasoning

Data quality and diversity are pivotal in constructing effective instruction-tuning datasets. With the increasing availability of open-source instruction-tuning datasets, it is advantageous to automatically select high-quality and diverse subsets from a vast amount of data. Existing methods typically prioritize instance quality and use heuristic rules to maintain diversity. However, this absence of a comprehensive view of the entire collection often leads to suboptimal results. Moreover, heuristic rules generally focus on distance or clustering within the embedding space, which fails to capture the intent of complex instructions in semantic space accurately. To bridge this gap, we propose a unified dataset information measurement method. This method models the semantic space by constructing a label graph and quantifies diversity based on the distribution of information within the graph. Based on such measurement, we further introduce an efficient sampling method that selects data samples iteratively to Maximize the Information Gain (MIG) in semantic space. Experiments on various datasets and base models demonstrate that MIG consistently outperforms state-of-the-art methods. Notably, the model fine-tuned with 5% Tulu3 data sampled by MIG achieves comparable performance to the official SFT model trained on the full dataset, with improvements of +5.73% on AlpacaEval and +6.89% on Wildbench. This finding shows the potential for unified dataset measurement in guiding instruction data selection. Code will be available.

MIG: Automatic Data Selection for Instruction Tuning by Maximizing Information Gain in Semantic Space

Automatic Term Extraction (ATE) plays a crucial role in identifying domain-specific terms for applications such as machine translation and information retrieval. While Large Language Models (LLMs) have significantly advanced various NLP tasks, their effectiveness in ATE remains largely unexplored. This work introduces a retrieval-based approach that prioritizes syntactic similarity over semantic similarity in few-shot selection, enhancing term boundary identification and annotation consistency. We assess its impact in both in-domain and cross-domain settings, examining the influence of term overlap between queries and retrieved examples. Experiments on ACTER, ACL RD-TEC 2.0, and BCGM demonstrate that our syntactic retrieval strategy improves ATE performance, particularly in low-resource and cross-domain scenarios.

Enhancing Automatic Term Extraction in Large Language Models via Syntactic Retrieval

In-Context Reinforcement Learning (ICRL) is a frontier paradigm to solve Reinforcement Learning (RL) problems in the foundation model era. While ICRL capabilities have been demonstrated in transformers through task-specific training, the potential of Large Language Models (LLMs) out-of-the-box remains largely un-explored. This paper investigates whether LLMs can generalize cross-domain to perform ICRL under the problem of Dueling Bandits (DB), a stateless preference-based RL setting. We reveal that our top-performing LLM exhibits notable zero-shot ability in relative decision-making, achieving low short-term weak regret across all the DB environment instances by quickly including the best arm in duels. However, an optimality gap exists between LLMs and classic DB algorithms in terms of strong regret. LLMs struggle to converge and consistently exploit even when explicitly prompted to do so, and are sensitive to prompt variations. To bridge this gap, we propose an agentic flow framework: LLM with Enhanced Algorithmic Dueling (LEAD), which integrates off-the-shelf DB algorithm support with LLM agents through fine-grained adaptive interplay. We show that LEAD has theoretical guarantees inherited from classic DB algorithms on both weak and strong regret. We validate its efficacy and robustness even with noisy and adversarial prompts. The design of such an agentic framework sheds light on how to enhance the trustworthiness of general-purpose LLMs generalized to in-context decision-making tasks.

Beyond Numeric Rewards: In-Context Dueling Bandits with LLM Agents

Large language models (LLMs) exhibit strong reasoning abilities, often attributed to few-shot or zero-shot chain-of-thought (CoT) prompting. While effective, these methods require labor-intensive prompt engineering, raising the question of whether reasoning can be induced without reliance on explicit prompts. In this work, we unlock the reasoning capabilities of LLMs without explicit prompting. Inspired by zero-shot CoT and CoT-decoding, we propose a novel decoding strategy that systematically nudges LLMs to continue reasoning, thereby preventing immature reasoning processes. Specifically, we monitor the model’s generation and inject a designated phrase whenever it is likely to conclude its response prematurely, before completing the reasoning process. Our experimental evaluations on diverse reasoning benchmarks demonstrate that our proposed strategy substantially improves LLM reasoning capabilities, highlighting the potential of decoding-based interventions as an alternative to traditional prompting techniques.

“Well, Keep Thinking”: Enhancing LLM Reasoning with Adaptive Injection Decoding

Knowledge Graphs (KGs) have gained popularity as a means of storing structured data, with property graphs, in particular, gaining traction in recent years. Consequently, the task of semantic parsing remains crucial in enabling access to the information in these graphs via natural language queries. However, annotated data is scarce, requires significant effort to create, and is not easily transferable between different graphs. To address these challenges we introduce SPOT, a method to generate training data for semantic parsing over Property Graphs without human intervention. We generate tree patterns, match them to the KG to obtain a query program, and use a finite-state transducer to produce a proto-natural language realization of the query. Finally, we paraphrase the proto-NL with an LLM to generate samples for training a semantic parser. We demonstrate the effectiveness of SPOT on two property graph benchmarks utilizing the Cypher query language. In addition, we show that our approach can also be applied effectively to RDF graphs.

SPOT: Zero-Shot Semantic Parsing Over Property Graphs

Recent studies on reasoning in language models (LMs) have sparked a debate on whether they can learn systematic inferential principles or merely exploit superficial patterns in the training data. To understand and uncover the mechanisms adopted for formal reasoning in LMs, this paper presents a mechanistic interpretation of syllogistic inference. Specifically, we present a methodology for circuit discovery aimed at interpreting content-independent and formal reasoning mechanisms. Through two distinct intervention methods, we uncover a sufficient and necessary circuit involving middle-term suppression that elucidates how LMs transfer information to derive valid conclusions from premises.
Furthermore, we investigate how belief biases manifest in syllogistic inference, finding evidence of partial contamination from additional attention heads responsible for encoding commonsense and contextualized knowledge. Finally, we explore the generalization of the discovered mechanisms across various syllogistic schemes, model sizes and architectures. The identified circuit is sufficient and necessary for syllogistic schemes on which the models achieve high accuracy ($\geq$ 60\%), with compatible activation patterns across models of different families. Overall, our findings suggest that LMs learn transferable content-independent reasoning mechanisms, but that, at the same time, such mechanisms do not involve generalizable and abstract logical primitives, being susceptible to contamination by the same world knowledge acquired during pre-training.

Reasoning Circuits in Language Models: A Mechanistic Interpretation of Syllogistic Inference

Large Language Models have emerged to show remarkable capabilities in the NLP domain. The effectiveness can mainly be attributed to the ability to adapt to an array of downstream tasks. However, generally, full fine-tuning is a computationally expensive job. To mitigate this, many techniques have been developed that prime efficiency, a prominent one being Low-Rank Adaptation (LoRA). However, LoRA and its variants employ re-parametrized additive updates. In this paper, we propose Low Rank Multiplicative Adaptation (LoRMA), which shifts the paradigm of additive updates to a much richer space of matrix multiplicative transformations. We tackle challenges such as computational complexity and rank inhibition by strategically ordering matrix operations and introducing rank inflation strategies. We conduct extensive experiments to show the effectiveness of our approach in terms of evaluation metrics and computational costs.

LoRMA: Low Rank Multiplicative Adaptation for LLMs

As Large Language Models are increasingly deployed in high-stakes domains, their ability to detect false assumptions and reason critically is crucial for ensuring reliable outputs. False-premise questions (FPQs) serve as an important evaluation method by exposing cases where flawed assumptions lead to incorrect responses. While existing benchmarks focus on single-hop FPQs, real-world reasoning often requires multi-hop inference, where models must verify consistency across multiple reasoning steps rather than relying on surface-level cues. To address this gap, we introduce MultiHoax, a benchmark for evaluating LLMs' ability to handle false premises in complex, multi-step reasoning tasks. Our dataset spans seven countries and ten diverse knowledge categories, using Wikipedia as the primary knowledge source to enable cross-regional factual reasoning. Experiments reveal that state-of-the-art LLMs struggle to detect false premises across different countries, knowledge categories, and multi-hop reasoning types, highlighting the need for improved false premise detection and more robust multi-hop reasoning capabilities in LLMs.

Downloads

Next from ACL 2025

Akan Cinematic Emotions (ACE): A Multimodal Multi-party Dataset for Emotion Recognition in Movie Dialogues

Similar lecture

KazMMLU: Evaluating Language Models on Kazakh, Russian, and Regional Knowledge of Kazakhstan