Austria

It is crucial for large language models (LLMs) to follow instructions that involve multiple constraints. In real-world scenarios, user instructions often contain soft constraints, which are semantically related and cannot be rule-based verified, posing challenges for LLMs. To enhance the soft constraint following ability of LLMs, we initially design a pipeline to construct datasets with high-quality outputs for instructions containing soft constraints automatically. Additionally, to fully utilize the positive and negative samples generated during the data construction process, we choose Direct Preference Optimization (DPO) as the training method. Furthermore, taking into account the difficulty of soft constraints indicated by the number of constraints, we design a curriculum learning training paradigm based on the constraint quantity. We experimentally evaluate the effectiveness of our methods in improving LLMs&#39; soft constraint following ability and analyze the factors driving the improvements.

ACL 2025

Step-by-Step Mastery: Enhancing Soft Constraint Following Ability of Large Language Models

soft constraint

curriculum learning

reinforcement learning

### 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)

It is crucial for large language models (LLMs) to follow instructions that involve multiple constraints. In real-world scenarios, user instructions often contain soft constraints, which are semantically related and cannot be rule-based verified, posing challenges for LLMs. To enhance the soft constraint following ability of LLMs, we initially design a pipeline to construct datasets with high-quality outputs for instructions containing soft constraints automatically. Additionally, to fully utilize the positive and negative samples generated during the data construction process, we choose Direct Preference Optimization (DPO) as the training method. Furthermore, taking into account the difficulty of soft constraints indicated by the number of constraints, we design a curriculum learning training paradigm based on the constraint quantity. We experimentally evaluate the effectiveness of our methods in improving LLMs' soft constraint following ability and analyze the factors driving the improvements.

poster

Current Multimodal Large Language Models (MLLMs) excel in general visual reasoning but remain underexplored in Abstract Visual Reasoning (AVR), which demands higher-order reasoning to identify abstract rules beyond simple perception. Existing AVR benchmarks focus on single-step reasoning, emphasizing the end result but neglecting the multi-stage nature of reasoning process. Past studies found MLLMs struggle with these benchmarks, but it doesn't explain how they fail. To address this gap, we introduce MultiStAR, a Multi-Stage AVR benchmark, based on RAVEN, designed to assess reasoning across varying levels of complexity. Additionally, existing metrics like accuracy only focus on the final outcomes while do not account for the correctness of intermediate steps. Therefore, we propose a novel metric, MSEval, which considers the correctness of intermediate steps in addition to the final outcomes. We conduct comprehensive experiments on MultiStAR using 17 representative close-source and open-source MLLMs. The results reveal that while existing MLLMs perform adequately on basic perception tasks, they continue to face challenges in more complex rule detection stages. The dataset and code will be available after acceptance.

Beyond Perception: Evaluating Abstract Visual Reasoning through Multi-Stage Task

Despite the remarkable success of transformer-based large language models (LLMs) across various domains, understanding and enhancing their mathematical capabilities remains a significant challenge. In this paper, we conduct a rigorous theoretical analysis of LLMs' mathematical abilities, with a specific focus on their arithmetic performances. We identify numerical precision as a key factor that influences their effectiveness in arithmetical tasks. Our results show that Transformers operating with low numerical precision fail to address arithmetic tasks, such as iterated addition and integer multiplication, unless the model size grows super-polynomially with respect to the input length. In contrast, Transformers with standard numerical precision can efficiently handle these tasks with significantly smaller model sizes. We further support our theoretical findings through empirical experiments that explore the impact of varying numerical precision on arithmetic tasks, providing valuable insights for improving the mathematical reasoning capabilities of LLMs.

How Numerical Precision Affects Arithmetical Reasoning Capabilities of LLMs

Understanding alignment techniques begins with comprehending zero-shot generalization brought by instruction tuning, but little of the mechanism has been understood. Existing work has largely been confined to the task level, without considering that tasks are artificially defined and, to LLMs, merely consist of tokens and representations. To bridge this gap, we investigate zero-shot generalization from the perspective of the data itself. We first demonstrate that zero-shot generalization happens very early during instruction tuning, with loss serving as a stable indicator. Next, we investigate training data arrangement through similarity and granularity perspectives, confirming that the timing of exposure to certain training examples may greatly facilitate generalization on unseen tasks. Finally, we propose a more grounded training data arrangement framework, Test-centric Multi-turn Arrangement, and show its effectiveness in promoting continual learning and further loss reduction. For the first time, we show that zero-shot generalization during instruction tuning is a form of similarity-based generalization between training and test data at the instance level.

The Right Time Matters: Data Arrangement Affects Zero-Shot Generalization in Instruction Tuning

We propose a training-free framework that enables large language models (LLMs) to effectively process long texts, using a divide-and-conquer strategy for comprehensive document understanding.The proposed LLM×MapReduce framework splits the entire document into several chunks for LLMs to read and then aggregates the intermediate outputs to produce the final response. The main challenge for divide-and-conquer long text processing frameworks lies in the risk of losing essential long-range information due to document splitting, which can lead the model to produce incomplete or incorrect answers based on the segmented texts.Disrupted long-range information can be classified into two categories: inter-chunk dependency and inter-chunk conflict.We design a structured information protocol to better cope with inter-chunk dependency and an in-context confidence calibration mechanism to resolve inter-chunk conflicts. Experiments demonstrate that LLM×MapReduce outperforms representative open-source and commercial long-context LLMs and is compatible with several models.Our framework can also function as a data synthesis engine, capable of generating high-quality long-alignment data using only short-context LLMs.

MFinMeeting: A Multilingual, Multi-Sector, and Multi-Task Financial Meeting Understanding Evaluation Dataset

Data Augmentation (DA) has emerged as a promising solution to address the scarcity of high-quality annotated data in low-resource relation extraction (LRE). Leveraging large language models (LLMs), DA has significantly improved the performance of RE models with considerably fewer parameters. However, existing
DA methods struggle with diversity misalignments, as they neglect the diversity required by the model and generate homogeneous augmentations that do not cover the inter-sample and inter-relation variability, leading to suboptimal performance. Inspired by the Observe-Orient-Decide-Act (OODA) framework, which provides a robust theoretical foundation for iterative decision-making under dynamic conditions, we propose an OODA-driven Diverse DA method (ODDA), guiding the data generation and selection process. ODDA first observes the RE model’s behavior to select effective demonstrations for LLMs. Next, it orients LLMs
towards generating diverse data by replacing schema constraints with attribute constraints. Then ODDA decides on the final augmented dataset with overall diversity from a global search and finally acts to train the RE model. Extensive experiments on three widely-used benchmarks demonstrate that ODDA consistently outperforms state-of-the-art baselines, achieving average F1 improvements of 3.1% across various LRE scenarios while maintaining enhanced model stability

ODDA: An OODA-Driven Diverse Data Augmentation Framework for Low-Resource Relation Extraction

Video summarization aims to generate a condensed textual version of an original video. Summaries may consist of either plain text or a shortlist of salient events, possibly including temporal or spatial references. Video Large Language Models (VLLMs) exhibit impressive zero-shot capabilities in video analysis. However, their performance varies significantly according to the LLM prompt, the characteristics of the video, and the properties of the training data and LLM architecture.In this work, we thoroughly evaluate the zero-shot summarization performance of four state-of-the-art open-source VLLMs specifically designed to address spatial and temporal reasoning. In light of the detected summarization issues, we propose different cost-effective mitigation strategies, based on Chain-of-Thought prompting, that involve the injection of knowledge extracted by external, lightweight models. To perform the VLLM evaluation, we design a new video summarization benchmark consisting of 100 videos with varying characteristics in terms of domain, duration, and spatio-temporal properties. Videos are manually annotated by three independent human experts with plain text, event-based, and spatio-temporal summaries. The experimental evaluation shows that VLLMs significantly benefit from prompting a list of recognized actions, whereas injecting automatically recognized objects and scene changes respectively improve spatially contextualized and event-based summaries in specific cases.

Detecting and Mitigating Challenges in Zero-Shot Video Summarization with Video LLMs

We introduce a novel multilingual and hierarchical corpus annotated for entity framing and role portrayal in news articles. The dataset uses a unique taxonomy inspired by storytelling elements, comprising 22 fine-grained roles, or archetypes, nested within three main categories: protagonist, antagonist, and innocent. Each archetype is carefully defined, capturing nuanced portrayals of entities such as guardian, martyr, and underdog for protagonists; tyrant, deceiver, and bigot for antagonists; and victim, scapegoat, and exploited for innocents. The dataset includes 1,378 recent news articles in five languages (Bulgarian, English, Hindi, European Portuguese, and Russian) focusing on two critical domains of global significance: the Ukraine-Russia War and Climate Change. Over 5,800 entity mentions have been annotated with role labels. This dataset serves as a valuable resource for research into role portrayal and has broader implications for news analysis. We describe the characteristics of the dataset and the annotation process, and we report evaluation results on fine-tuned state-of-the-art multilingual transformers and hierarchical zero-shot learning using LLMs at the level of a document, a paragraph, and a sentence.

Entity Framing and Role Portrayal in the News

Large Language Models (LLMs) have shown impressive reasoning capabilities, yet existing prompting methods face a critical trade-off: simple approaches often struggle with complex tasks and reasoning stability, while more sophisticated methods require multiple inferences and substantial computational resources, limiting their practical deployment. To address this challenge, we propose Derailer-Rerailer, a novel framework that adaptively balances reasoning accuracy and computational efficiency. At its core, our framework employs a lightweight Derailer mechanism to assess reasoning stability and selectively triggers an advanced Rerailer verification process only when necessary, thereby optimizing computational resource usage. Extensive evaluation across both open and closed-source models on more than 20 categories of mathematical, symbolic, and commonsense reasoning tasks demonstrates our framework's effectiveness: Derailer-Rerailer achieves significant accuracy improvements (8-11% across various reasoning tasks) while maintaining 2-3 times better efficiency than existing verification methods, with particularly strong performance in mathematical and symbolic reasoning, offering a practical solution for enhancing LLM reasoning reliability while significantly reducing computational overhead.

Derailer-Rerailer: Adaptive Verification for Efficient and Reliable Language Model Reasoning

When confronting long document information retrieval for Query-Focused Summarization(QFS), Traditional Retrieval-Augmented Generation(RAG) frameworks struggle to retrieve all relevant knowledge points, and the chunking and retrieve strategies of existing frameworks may disrupt the connections between knowledge points and the integrity of the information. To address these issues, we propose TreeRAG, which employs Tree-Chunking for chunking and embedding in a tree-like structure , coupled with "root-to-leaves" and "leaf-to-root" retrieve strategy named Bidirectional Traversal Retrieval. This approach effectively preserves the hierarchical structure among knowledge points and significantly enhances the ability to retrieve while minimizing noise inference. Our experimental results on the Finance, Law, and Medical subsets of the Dragonball dataset demonstrate that TreeRAG achieves significant enhancements in both recall quality and precision compared to traditional and popular existing methods and achieves better performance to corresponding question-answering tasks, marking a new breakthrough in long document knowledge retrieval.

TreeRAG: Unleashing the Power of Hierarchical Storage for Enhanced Knowledge Retrieval in Long Documents

To assist humans in efficiently validating RAG-generated content, developing a fine-grained attribution mechanism that provides supporting evidence from retrieved documents for every answer span is essential. Existing fine-grained attribution methods rely on model-internal similarity metrics between responses and documents, such as saliency scores and hidden state similarity. However, these approaches suffer from either high computational complexity or coarse-grained representations. Additionally, a common problem shared by the previous works is their reliance on decoder-only Transformers, limiting their ability to incorporate contextual information after the target span. To address the above problems, we propose two techniques applicable to all model-internals-based methods. First, we aggregate token-wise evidence through set union operations, preserving the granularity of representations. Second, we enhance the attributor by integrating dependency parsing to enrich the semantic completeness of target spans. For practical implementation, our approach employs attention weights as the similarity metric. Experimental results demonstrate that the proposed method consistently outperforms all prior works.

Downloads

Next from ACL 2025

Beyond Perception: Evaluating Abstract Visual Reasoning through Multi-Stage Task

Similar lecture

CARE-STaR: Constraint-aware Self-taught Reasoner