Austria

Misinformation about climate science is a serious challenge for our society. This paper introduces CPIQA (Climate Paper Image Question-Answering), a new question-answer dataset featuring 4,551 full-text open-source academic papers in the area of climate science with 54,612 GPT-4o generated question-answer pairs. CPIQA contains four question types (numeric, figure-based, non-figure-based, reasoning), each generated using three user roles (expert, non-expert, climate sceptic). CPIQA is multimodal, incorporating information from figures and graphs with GPT-4o descriptive annotations. We describe Context-RAG, a novel method for RAG prompt decomposition and augmentation involving extracting distinct contexts for the question. Evaluation results for Context-RAG on the benchmark SPIQA dataset outperforms the previous best state of the art model in two out of three test cases. For our CPIQA dataset, Context-RAG outperforms our standard RAG baseline on all five base LLMs we tested, showing our novel contextual decomposition method can generalize to any LLM architecture. Expert evaluation of our best performing model (GPT-4o with Context-RAG) by climate science experts highlights strengths in precision and provenance tracking, particularly for figure-based and reasoning questions.

ACL 2025

CPIQA: Climate Paper Image Question Answering Dataset for Retrieval-Augmented Generation with Context-based Query Expansion

### 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)

workshop paper

Large language models that use retrieval augmented generation have the potential to unlock valuable knowledge for researchers, policymakers, and the public by making long and technical climate-related documents more accessible. While this approach can help alleviate factual hallucinations by relying on retrieved passages as additional context, its effectiveness depends on whether the model's output remains faithful to these passages. To address this, we explore the automatic assessment of faithfulness of different models in this setting. We then focus on ClimateGPT, a large language model specialised in climate science, to examine which factors in its instruction fine-tuning impact the model's faithfulness. By excluding unfaithful subsets of the model's training data, we develop ClimateGPT Faithful+, which achieves an improvement in faithfulness from 30\% to 57\% in supported atomic claims according to our automatic metric.

Listen to the Context: Towards Faithful Large Language Models for Retrieval Augmented Generation on Climate Questions

Research syntheses, such as systematic maps or evidence and gap maps, provide valuable overviews of the coverage of research in a particular field.
They serve as pointers for funders and researchers to identify important gaps in the literature where more research is needed but also to find relevant work for more in-depth systematic reviews or meta-analyses.
However, systematic maps become outdated quickly, sometimes even after they are released due to the time it takes to screen and code the available literature and long publication processes.
Furthermore, the write-up of the synthesis (in form of a peer-reviewed article) can only serve as a high-level summary---for detailed questions one would need full access to the underlying data.
To this end, we developed an interactive web-based platform to share annotated datasets.
For some datasets, where automated categorisation passes the necessary scientific quality standards, we also update the data as new research becomes available and thus make them `living'.

Interactive platform for the exploration of large-scale 'living' systematic maps

Accurately synthesizing climate evidence into concise statements is crucial for policy making and fostering public trust in climate science. Recent advancements in Large Language Models (LLMs), particularly the emergence of reasoning-optimized variants, which excel at mathematical and logical tasks, present a promising yet untested opportunity for scientific evidence synthesis. We evaluate state-of-the-art reasoning LLMs on two key tasks: (1) *contextual confidence classification*, assigning appropriate confidence levels to climate statements based on evidence, and (2) *factual summarization of climate evidence*, generating concise summaries evaluated for coherence, faithfulness, and similarity to expert-written versions. Using a novel dataset of 612 structured examples constructed from the Sixth Assessment Report (AR6) of the Intergovernmental Panel on Climate Change (IPCC), we find reasoning LLMs outperform general-purpose models in confidence classification by 8 percentage points in accuracy and macro-F1 scores. However, for summarization tasks, performance differences between model types are mixed. Our findings demonstrate that reasoning LLMs show promise as auxiliary tools for confidence assessment in climate evidence synthesis, while highlighting significant limitations in their direct application to climate evidence summarization. This work establishes a foundation for future research on the targeted integration of LLMs into scientific assessment workflows.

Can Reasoning LLMs Synthesize Complex Climate Statements?

The rapid spread of multilingual misinformation requires robust automated fact verification systems capable of handling fine-grained veracity assessments across diverse languages. While large language models have shown remarkable capabilities across many NLP tasks, their effectiveness for multilingual claim verification with nuanced classification schemes remains understudied. We conduct a comprehensive evaluation of five state-of-the-art language models on the X-Fact dataset, which spans 25 languages with seven distinct veracity categories. Our experiments compare small language models (encoder-based XLM-R and mT5) with recent decoder-only LLMs (Llama 3.1, Qwen 2.5, Mistral Nemo) using both prompting and fine-tuning approaches. Surprisingly, we find that XLM-R (270M parameters) substantially outperforms all tested LLMs (7-12B parameters), achieving 57.7% macro-F1 compared to the best LLM performance of 16.9%. This represents a 15.8% improvement over the previous state-of-the-art (41.9%), establishing new performance benchmarks for multilingual fact verification. Our analysis reveals problematic patterns in LLM behavior, including systematic difficulties in leveraging evidence and pronounced biases toward frequent categories in imbalanced data settings. These findings suggest that for fine-grained multilingual fact verification, smaller specialized models may be more effective than general-purpose large models, with important implications for practical deployment of fact-checking systems.

When Scale Meets Diversity: Evaluating Language Models on Fine-Grained Multilingual Claim Verification

Retrieval-augmented generation (RAG) helps mitigate hallucinations in large language models (LLMs). However, LLMs struggle to maintain consistent reasoning when exposed to misleading evidence, especially in the political domain, where information is polarized or selectively framed. Many existing RAG benchmarks assume a clean retrieval setting while noisy RAG benchmarks lack naturally misleading evidence, overestimating real-world performance. We introduce \textsc{RAGuard}, a fact-checking dataset designed to evaluate the robustness of RAG systems against misleading retrievals. Unlike prior benchmarks that rely on synthetic noise, our dataset constructs its retrieval corpus from Reddit discussions, capturing naturally occurring misinformation categorized as \textit{supporting}, \textit{misleading}, and \textit{unrelated}. Our experiments reveal that, when exposed to potentially misleading retrievals, all tested RAG systems perform worse than their zero-shot baselines (i.e., no retrieval at all), while human annotators perform similarly or better, highlighting LLMs' susceptibility to noisy environments. To our knowledge, \textsc{RAGuard} is the first benchmark to systematically assess the robustness of the RAG against misleading evidence. We expect this benchmark to drive future research toward improving RAG beyond idealized datasets, making them more reliable for real-world applications.

Worse than Zero-shot? A Fact-Checking Dataset for Evaluating the Robustness of RAG Against Misleading Retrievals

Social media has emerged as a valueable source for early pandemic detection, as repeated mentions of symptoms by users may signal the onset of an outbreak. However, to be a reliable system, validation through fact-checking and verification against official health records is essential. Without this step, systems risk spreading misinformation to the public. The effectiveness of these systems also depend on their ability to process data in multiple languages, given the multilingual nature of social media data. Yet, many NLP datasets and disease surveillance system remain heavily English-centric, leading to significant performance gaps for low-resource languages. This issue is especially critical in Southeast Asia, where symptom expression may vary culturally and linguistically. Therefore, this study evaluates the symptom detection capabilities of LLMs in social media posts across multiple languages, models, and symptoms to enhance health-related fact-checking. Our results reveal significant language-based discrepancies, with European languages outperforming under-resourced Southeast Asian languages. Furthermore, we identify symptom-specific challenges, particularly in detecting respiratory illnesses such as influenza, which LLMs tend to overpredict. The overestimation or misclassification of symptom mentions can lead to false alarms or public misinformation when deployed in real-world settings. This underscores the importance of symptom detection as a critical first step in medical fact-checking within early outbreak detection systems.



Multilingual Symptom Detection on Social Media: Enhancing Health-related Fact-checking with LLMs


We propose a method of improving the performance of question answering based on the interpretation of criminal law regulations in the Korean language by using large language models. In this study, we develop a system that accumulates legislative texts and case precedents related to criminal procedures published on the Internet.The system searches for relevant legal provisions and precedents related to the query under the RAG (Retrieval-Augmented Generation) framework. It generates accurate responses to questions by conducting reasoning through large language models based on these relevant laws and precedents. As an application example of this system, it can be utilized to support decision making in investigations and legal interpretation scenarios within the field of Korean criminal law.

RAG based Question Answering of Korean Laws and Precedents

Structured fact verification benchmarks like AVeriTeC decompose claims into QA pairs to support fine-grained reasoning. However, current systems generate QA pairs independently for each evidence sentence, leading to redundancy, drift, and noise. We introduce a modular LLM-based QA consolidation module that jointly filters, clusters, and rewrites QA pairs at the claim level. Experiments show that this method improves evidence quality and veracity prediction accuracy. Our analysis also highlights the impact of model scale and alignment on downstream performance.


GQC: LLM-Based Grouped QA Consolidation for Open-Domain Fact Verification at AVeriTeC

While Large Language Models (LLMs) with retrieval augmented generation (RAG) capabilities are increasingly used to verify information, their reliability for political fact-checking remains questionable. We present a comprehensive evaluation of eight state-of-the-art LLMs on their ability to fact-check political statements related to the 2024 U.S. presidential election. Using a dataset of 1,374 statements derived from 530 original claims fact-checked by major organizations in the run-up to the elections, we test both original and reformulated versions of statements. Our findings reveal that even the best-performing models achieve only modest accuracy (macro F1 score of 0.51), with RAG providing minimal improvements over models without search capabilities. Models particularly struggle with nuanced "misleading" statements and demonstrate poor robustness to reformulations. These results indicate that users relying on LLMs for political fact-checking are likely to receive inconsistent and sometimes incorrect assessments, even for statements that have been previously fact-checked by professional organizations and are readily available online.

RAG Is Not Enough: Evidence of Fact-Checking Limitations During the 2024 U.S. Presidential Elections

Automated fact-checking (AFC) of factual claims require efficiency and accuracy. Existing evaluation frameworks like Ev2R achieve strong semantic grounding but incur substantial computational cost, while simpler metrics based on overlap or one-to-one matching often misalign with human judgments. In this paper, we introduce SemQA, a lightweight and accurate evidence-scoring metric that combines transformer-based question scoring with bidirectional NLI entailment on answers. We evaluate SemQA by conducting human evaluations, analyzing correlations with existing metrics, and examining representative examples.

Downloads

Next from ACL 2025

Listen to the Context: Towards Faithful Large Language Models for Retrieval Augmented Generation on Climate Questions

Similar lecture

Question answering in Climate Adaptation: Model Development and Evaluation with Expert Feedback