Austria

Recently, Large Language Models (LLMs) have demonstrated significant potential for data annotation, markedly reducing the labor costs associated with downstream applications. However, existing methods mostly adopt an aggressive strategy by prompting LLM to determine a single gold label for each unlabeled sample. Due to the inherent uncertainty within LLMs, they often produce incorrect labels for difficult samples, severely compromising the data quality for downstream applications. Motivated by ambiguity aversion in human behaviors, we propose a novel candidate annotation paradigm wherein large language models are encouraged to output all possible labels when incurring uncertainty. To ensure unique labels are provided for downstream tasks, we develop a teacher-student framework CanDist that distills candidate annotations with a Small Language Model (SLM). We further provide a rigorous justification demonstrating that distilling candidate annotations from the teacher LLM offers superior theoretical guarantees compared to directly using single annotations. Extensive experiments across six text classification tasks validate the effectiveness of our proposed method. The source code is available at https://github.com/MingxuanXia/CanDist.

ACL 2025

Prompt Candidates, then Distill: A Teacher-Student Framework for LLM-driven Data Annotation

data-efficient training

### 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

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)

Existing attacks against multimodal language models often communicate instruction through text, either as an explicit malicious instruction or a crafted generic prompt, and accompanied by a toxic image. In contrast, here we exploit the capabilities of MLLMs in following non-textual instruction, i.e., an adversarial image or audio, namely Con Instruction. It is a novel gray-box attack method that generates adversarial images or audio to convey specific harmful instructions to MLLMs. We also find that combining our adversarial examples with certain non-empty text inputs amplifies attack success, while appending these after malicious text has limited effects. To evaluate whether an attack is successful, we introduce a new attack response categorization (ARC) that considers the response quality and relevancy concerning the malicious instruction. The results show that Con Instruction effectively bypasses the safety mechanisms in various visual and audio-language models, including LLaVA-v1.5, InternVL, Qwen-VL, and Qwen-Audio, across two standard benchmarks: AdvBench and SafeBench. Specifically, our method achieves the highest attack success rates, reaching 81.3% and 86.6% on LLaVA-v1.5 (13B). We show that larger models are more susceptible toCon Instruction, contrasting observations in their underlying LLMs. On the defense side, we explore various methods against our attacks and find substantial gaps among existing techniques. The code will be made available upon publication.

Con Instruction: Universal Jailbreaking of Multimodal Large Language Models via Non-Textual Modalities

Dynamic Retrieval-augmented Generation (RAG) has shown great success in mitigating hallucinations in large language models (LLMs) during generation. However, existing dynamic RAG methods face significant limitations in two key aspects: 1) Lack of an effective mechanism to control retrieval triggers, and 2) Lack of effective scrutiny of retrieval content. To address these limitations, we propose an innovative dynamic RAG method, DioR (Adaptive Cognitive Detection and Contextual Retrieval Optimization), which consists of two main components: adaptive cognitive detection and contextual retrieval optimization, specifically designed to determine when retrieval is needed and what to retrieve for LLMs is useful. Experimental results demonstrate that DioR achieves superior performance on all tasks, demonstrating the effectiveness of our work.

DioR: Adaptive Cognitive Detection and Contextual Retrieval Optimization for Dynamic Retrieval-Augmented Generation

Maximum a posteriori decoding, a commonly used method for neural machine translation (NMT), aims to maximize the estimated posterior probability. However, high estimated probability does not always lead to high translation quality. Minimum Bayes Risk (MBR) decoding offers an alternative by seeking hypotheses with the highest expected utility.

Inspired by Quality Estimation (QE) reranking which uses the QE model as a ranker, we propose source-based MBR (sMBR) decoding, a novel approach that utilizes quasi-sources (generated via paraphrasing or back-translation) as ``support hypotheses'' and a reference-free quality estimation metric as the utility function, marking the first work to solely use sources in MBR decoding. Experiments show that sMBR outperforms QE reranking and the standard MBR decoding. Our findings suggest that sMBR is a promising approach for NMT decoding.

Unveiling the Power of Source: Source-based Minimum Bayes Risk Decoding for Neural Machine Translation

Embodied question answering (EQA) means using *perception of* and *action in* an environment to answer natural language questions about that environment. However, previous work has demonstrated that blind language models (which do not incorporate perception, but predict an answer based solely on the question text) are a strong baseline for existing benchmarks, even compared against state-of-the-art vision and language models. To determine whether a model is grounding its answers in its specific environment, rather than relying on a language model's expectations about the world generally, we propose PQB-EQA, a *per-question balanced* EQA dataset. In this new benchmark, every question appears twice, paired with two different environments that yield two different answers. That is, the answer distribution is balanced for each question, not just across the whole dataset. We show both theoretically and empirically that grounding in the environment is necessary to perform better than chance on PQB-EQA.

Grounded, or a Good Guesser? A Per-Question Balanced Dataset to Separate Blind from Grounded Models for Embodied Question Answering

Music performances, characterized by dense and continuous audio as well as seamless audio-visual integration, present unique challenges for multimodal scene understanding and reasoning. Recent Music Performance Audio-Visual Question Answering (Music AVQA) datasets have been proposed to reflect these challenges, highlighting the continued need for more effective integration of audio-visual representations in complex question answering. However, existing Music AVQA methods often rely on dense and unoptimized representations, leading to inefficiencies in the isolation of key information, the reduction of redundancy, and the prioritization of critical samples. To address these challenges, we introduce Sparsify, a sparse learning framework specifically designed for Music AVQA. It integrates three sparsification strategies into an end-to-end pipeline and achieves state-of-the-art performance on the Music AVQA datasets. In addition, it reduces training time by 28.32% compared to its fully trained dense counterpart while maintaining accuracy, demonstrating clear efficiency gains. To further improve data efficiency, we propose a key-subset selection algorithm that selects and uses approximately 25% of MUSIC-AVQA v2.0 training data and retains 70–80% of full-data performance across models.

Learning Sparsity for Effective and Efficient Music Performance Question Answering

Multi-step multimodal reasoning tasks pose significant challenges for multimodal large language models (MLLMs), and finding effective ways to enhance their performance in such scenarios remains an unresolved issue. In this paper, we propose AR-MCTS, a universal framework designed to progressively improve the reasoning capabilities of MLLMs through Active Retrieval (AR) and Monte Carlo Tree Search (MCTS). AR-MCTS follows the MCTS algorithm and heuristically integrates an active retrieval mechanism during the expansion stage to automatically acquire high-quality step-wise reasoning annotations. Moreover, we further introduce curriculum training objectives to progressively align with a process reward model, ultimately achieving process-level multimodal reasoning verification. Experimental results across three complex multimodal reasoning benchmarks confirm the effectiveness of AR-MCTS. Further analysis demonstrates that it can optimize sampling diversity and accuracy, yielding reliable multimodal reasoning.

Progressive Multimodal Reasoning via Active Retrieval

Large Vision-Language Models (LVLMs) have shown impressive progress by integrating visual perception with linguistic understanding to produce contextually grounded outputs. Despite these advancements achieved, LVLMs still suffer from the hallucination problem, e.g., they tend to produce content that does not exist in the input images. Our investigation suggests that such hallucinations often stem from the deficiencies in fine-grained comprehension on the visual aspect, particularly when visual scenes exhibit appearance or semantic similarities (e.g., bicycle vs. motorcycles, baseball bat vs. baseball). In this work, we show such hallucination is naturally mitigated via a novel method called visual evidence prompting, utilizing small visual models to complement the LVLMs. While traditional visual models are not adept at interacting with humans, they excel at perceiving the fine-grained image contents. By symbolizing the professional outputs of domain-expert models as prompts, the LVLM generalists are able to refer to these evidences as visual knowledge to generate more precise answers. Detailed analysis shows that visual evidence enables models to adjust and rectify the attribution and attention on the images, reducing visual confusion by suppressing false activation while enhancing correct ones. Extensive experiments and in-depth analysis demonstrate the effectiveness of our method. We hope our straightforward but insightful work enhances the comprehension of hallucination in LVLMs and offers valuable perspectives on addressing such challenges.

Visual Evidence Prompting Mitigates Hallucinations in Large Vision-Language Models

Process Reward Models (PRMs) are crucial for guiding Large Language Models (LLMs) in complex scenarios by providing dense reward signals. However, existing PRMs primarily rely on heuristic approaches, which struggle with cross-domain generalization. While LLM-as-judge has been proposed to provide generalized rewards, current research has focused mainly on feedback results, overlooking the meaningful guidance embedded within the text. Additionally, static and coarse-grained evaluation criteria struggle to adapt to complex process supervision. To tackle these challenges, we propose Dynamic and Generalizable Process Reward Modeling (DG-PRM), which features a reward tree to capture and store fine-grained, multi-dimensional reward criteria. DG-PRM dynamically selects reward signals for step-wise reward scoring. To handle multifaceted reward signals, we pioneeringly adopt Pareto dominance estimation to identify discriminative positive and negative pairs. Experimental results show that DG-PRM achieves stunning performance on prevailing benchmarks, significantly boosting model performance across tasks with dense rewards. Further analysis reveals that DG-PRM adapts well to out-of-distribution scenarios, demonstrating exceptional generalizability.

Dynamic and Generalizable Process Reward Modeling

Video dubbing aims to translate original speech in visual media programs from the source language to the target language, relying on neural machine translation and text-to-speech technologies. Due to varying information densities across languages, target speech often mismatches the source speech duration, causing audio-video synchronization issues that significantly impact viewer experience. In this study, we approach duration alignment in LLM-based video dubbing machine translation as a preference optimization problem. We propose the Segment Supervised Preference Optimization (SSPO) method, which employs a segment-wise sampling strategy and fine-grained loss to mitigate duration mismatches between source and target lines. Experimental results demonstrate that SSPO achieves superior performance in duration alignment tasks.

Fine-grained Video Dubbing Duration Alignment with Segment Supervised Preference Optimization

The advent of test-time scaling in large language models (LLMs), exemplified by OpenAI’s o1 series, has advanced reasoning capabilities by scaling computational resource allocation during inference. While successors like QwQ, Deepseek-R1 (R1) and LIMO replicate these advancements, whether these models truly possess test-time scaling capabilities remains underexplored. This study found that longer CoTs of these o1-like models do not consistently enhance accuracy; in fact, correct solutions are often shorter than incorrect ones for the same questions. Further investigation shows this phenomenon is closely related to models' self-revision capabilities - longer CoTs contain more self-revisions, which often lead to performance degradation. We then compare sequential and parallel scaling strategies on QwQ, R1 and LIMO, finding that parallel scaling achieves better coverage and scalability. Based on these insights, we propose ``Shortest Majority Vote'', a method that combines parallel scaling strategies with CoT length characteristics, significantly improving models' test-time scalability compared to conventional majority voting approaches.

Downloads

Next from ACL 2025

Con Instruction: Universal Jailbreaking of Multimodal Large Language Models via Non-Textual Modalities

Similar lecture

Learning with Less: Knowledge Distillation from Large Language Models via Unlabeled Data