Austria

Large language models (LLMs) have demonstrated remarkable reasoning capability in solving mathematical problems. However, existing approaches primarily focus on improving the quality of correct training data, e.g., distilling high-quality correct solutions from advanced models, neglecting the value contained in error data, potentially hindering the model’s reflective ability. Though some studies attempted to leverage error data, they often involve complex mechanisms, such as Monte Carlo Tree Search (MCTS) to explore error nodes.In this work, we propose to enhance LLM’s reasoning ability by Learning from Errors for MatheMatical Advancement (LEMMA). LEMMA constructs data consists of an incorrect solution with an erroneous step and a reflection connection to a correct solution for fine-tuning. Specifically, we systematically analyze the model-generated error types and introduce an _error-type grounded mistake augmentation_ method to collect diverse and representative errors. Correct solutions are either from fixing the errors or generating a fresh start. By fine-tuning on the constructed dataset, the model is able to _self-correct errors autonomously_ within the generation process _without relying on external critique models_. Experimental results demonstrate that LEMMA achieves significant performance improvements over other strong models with less than 90k data.

ACL 2025

LEMMA: Learning from Errors for MatheMatical Advancement in LLMs

methematical

self-correction

data synthesis

reflection

### 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

We introduce Physics, a comprehensive benchmark for PhD-qualifying exam physics problem solving. It contains 1297 expert-annotated problems covering six core areas: classical mechanics, quantum mechanics, thermodynamics and statistical mechanics, electromagnetism, atomic physics, and optics. Each problem requires advanced physics knowledge and mathematical reasoning. We develop a robust automated evaluation system for precise and reliable validation. Our evaluation of leading foundation models reveals substantial limitations. Even the most advanced model, o3-mini, achieves only 59.9% accuracy, highlighting significant challenges in solving high-level scientific problems. Through comprehensive error analysis, exploration of diverse prompting strategies, and RAG-based knowledge augmentation, we identify key areas for improvement, laying the foundation for future advancements.

Physics: Benchmarking Foundation Models for PhD-Qualifying Exam Physics Problem Solving

Evaluating Retrieval-Augmented Generation (RAG) in large language models (LLMs) is challenging because benchmarks can quickly become stale. Questions initially requiring retrieval may become answerable from pretraining knowledge as newer models incorporate more recent information during pretraining, making it difficult to distinguish evidence-based reasoning from recall. We introduce NeoQA (News Events for Out-of-training Question Answering), a benchmark designed to address this issue. To construct NeoQA, we generated timelines and knowledge bases of fictional news events and entities along with news articles and Q&A pairs to prevent LLMs from leveraging pretraining knowledge, ensuring that no prior evidence exists in their training data. We propose our dataset as a new platform for evaluating evidence-based question answering, as it requires LLMs to generate responses exclusively from retrieved evidence and only when sufficient evidence is available. NeoQA enables controlled evaluation across various evidence scenarios, including cases with missing or misleading details. Our findings indicate that LLMs struggle to distinguish subtle mismatches between questions and evidence, and suffer from short-cut reasoning when key information required to answer a question is missing from the evidence, underscoring key limitations in evidence-based reasoning.

NeoQA: Evidence-based Question Answering with Generated News Events

Current large language model (LLM) applications often employ multi-component prompts, comprising both system and user prompts, to guide model behaviors. While recent advancements have demonstrated the efficacy of automatically optimizing either the system or user prompt to boost performance, such unilateral approaches often yield suboptimal outcomes due to the interdependent nature of these components. In this work, we introduce P3, a novel self-improvement framework that concurrently optimizes both system and user prompts through an iterative process. The offline optimized prompts are further leveraged to promote online prompting by performing query-dependent prompt optimization. Extensive experiments on general tasks (e.g., Arena-hard and Alpaca-eval) and reasoning tasks (e.g., GSM8K and GPQA) demonstrate that P3 achieves superior performance in the realm of automatic prompt optimization. Our results highlight the effectiveness of a holistic optimization strategy in enhancing LLM performance across diverse domains.

P3: Prompts Promote Prompting

The main goal of this work is the creation of the Italian version of the WebNLG corpus through the application of Neural Machine Translation (NMT) and post-editing with hand-written rules. To achieve this goal, in a first step, several existing NMT models were analysed and compared in order to identify the system with the highest performance on the original corpus. In a second step, after using the best NMT system, we semi-automatically designed and applied a number of rules to refine and improve the quality of the produced resource, creating a new corpus named WebNLG-IT. We used this resource for fine-tuning several LLMs for RDF-to-text tasks. In this way, comparing the performance of LLM-based generators on both Italian and English, we have (1) evaluated the quality of WebNLG-IT with respect to the original English version, (2) released the first fine-tuned LLM-based system for generating Italian from semantic web triples and (3) introduced an Italian version of a modular generation pipeline for RDF-to-text.

WebNLG-IT: Construction of an aligned RDF-Italian corpus through Machine Translation techniques

We make the case for language models over logical forms (LFLMs), arguing that such models are more data-efficient than their textual counterparts. To that end, we introduce the Graph-based Formal-Logical Distributional Semantics (GFoLDS) prototype, a pretrained LM over graph representations of logical forms, as a proof-of-concept of LFLMs. Using GFoLDS, we present strong experimental evidence that LFLMs can leverage the built-in, basic linguistic knowledge inherent in such models to immediately begin learning more complex patterns. On downstream tasks, we show that GFoLDS vastly outperforms textual, transformer LMs (BERT) pretrained on the same data, indicating that LFLMs can learn with substantially less data than models over plain text. Furthermore, we show that the performance of this model is likely to scale with additional parameters and pretraining data, suggesting the viability of LFLMs in real-world applications.

Exploring Graph Representations of Logical Forms for Language Modeling

With the rapid emergence of novel capabilities
in Large Language Models (LLMs), the need
for rigorous multilingual and multicultural
benchmarks that are integrated has become
more pronounced. Though existing LLM
benchmarks are capable of evaluating specific
capabilities of LLMs in English as well as
in various mid- to low-resource languages,
including those in the Southeast Asian
(SEA) region, a comprehensive and authentic
evaluation suite for the SEA languages has
not been developed thus far. Here, we
present SEA-HELM, a holistic linguistic
and cultural LLM evaluation suite that
emphasizes SEA languages, comprising five
core pillars: (1) NLP CLASSICS, (2)
LLM-SPECIFICS, (3) SEA LINGUISTICS, (4)
SEA CULTURE, (5) SAFETY. SEA-HELM
currently supports Filipino, Indonesian, Tamil,
Thai, and Vietnamese. We also introduce
the SEA-HELM leaderboard, which allows
users to understand models’ multilingual and
multicultural performance in a systematic and
user-friendly manner.

SEA-HELM: Southeast Asian Holistic Evaluation of Language Models

Direct Preference Optimization (DPO) is broadly utilized for aligning Large Language Models (LLMs) with human values because of its flexibility. Despite its effectiveness, it has been observed that the capability of DPO to generate human-preferred response is limited and the results of DPO are far from resilient. To address these limitations, in this paper we propose a novel Self-Guided Direct Preference Optimization algorithm, i.e., SGDPO, which incorporates a pilot term to steer the gradient flow during the optimization process, allowing for fine-grained control over the updates of chosen and rejected rewards. We provide a detailed theoretical analysis of our proposed method and elucidate its operational mechanism. Furthermore, we conduct comprehensive experiments on various models and benchmarks. The extensive experimental results demonstrate the consistency between the empirical results and our theoretical analysis and confirm the effectiveness of our proposed approach (up to 9.19% higher score).

SGDPO: Self-Guided Direct Preference Optimization for Language Model Alignment

The widespread use of Large Multimodal Models (LMMs) has raised concerns about model toxicity. However, current research mainly focuses on explicit toxicity, with less attention to some more implicit toxicity regarding prejudice and discrimination. To address this limitation, we introduce a subtler type of toxicity named dual-implicit toxicity and a novel toxicity benchmark termed MDIT-Bench: Multimodal Dual-Implicit Toxicity Benchmark. Specifically, we first create the MDIT-Dataset with dual-implicit toxicity using the proposed Multi-stage Human-in-loop In-context Generation method. Based on this dataset, we construct the MDIT-Bench, a benchmark for evaluating the sensitivity of models to dual-implicit toxicity, with 317,638 questions covering 12 categories, 23 subcategories, and 780 topics. MDIT-Bench includes three difficulty levels, and we propose a metric to measure the toxicity gap exhibited by the model across them. In the experiment, we conducted MDIT-Bench on 13 prominent LMMs, and the results show that these LMMs cannot handle dual-implicit toxicity effectively. The model's performance drops significantly in hard level, revealing that these LMMs still contain a significant amount of hidden but activatable toxicity. The data will be released upon the paper's acceptance.

MDIT-Bench: Evaluating the Dual-Implicit Toxicity in Large Multimodal Models

Routing networks in sparse activated mixture-of-experts dynamically allocate input tokens to top-k experts through differentiable sparse transformations, which can expand neural networks while maintaining computational efficiency. Traditional MoE networks use an expert capacity constraint to achieve GPU-friendly computing. However, when the token is routed to experts with saturated capacity, token dropping occurs, and the padding operation of underutilized experts leads to low hardware efficiency. Removing the capacity constraint can also result in low computational efficiency and decreased load balancing. We propose **Maximum Score Routing (MaxScore)** in this work, a new MoE routing paradigm developed by modeling MoE routing as the minimum cost maximum flow problem and integrating a Soft-Topk operator. MaxScore retains the expert capacity constraint, and under the same FLOPs, compared to the constrained and unconstrained baselines, MaxScore exhibits lower training losses and higher evaluation scores.

Maximum Score Routing For Mixture-of-Experts

Pre-trained language models have shown remarkable performances in recent years, setting a new paradigm for natural language processing (NLP) research. The legal domain has received some attention from the NLP community partly due to its textual nature. Question-answering (QA) systems represent some of the tasks in this domain. This work explores the legal Multiple-Choice QA (MCQA) for a low-resource language, Romanian. The contribution of this work is multi-fold. We introduce JuRO, the first openly available Romanian legal MCQA dataset, comprising 10,836 questions from three examinations. Along with this dataset, we introduce CROL, an organized corpus of laws with a total of 93 distinct documents with their modifications over 763 time spans, which we used for information retrieval techniques in this work. Additionally, we construct Law-RoG, the first graph of legal knowledge for the Romanian language derived from the corpus mentioned above. Lastly, we propose a novel approach for MCQA, namely Graph Retrieval Augmented by Facts (GRAF), which achieves competitive results with generally accepted state-of-the-art methods and even exceeds them in most settings.

Downloads

Next from ACL 2025

Physics: Benchmarking Foundation Models for PhD-Qualifying Exam Physics Problem Solving

Similar lecture

Subtle Errors in Reasoning: Preference Learning via Error-injected Self-editing