Natural language explanations (NLEs) are vital for elucidating the reasoning behind large language model (LLM) decisions. Many techniques have been developed to generate NLEs using LLMs. However, like humans, LLMs might not always produce optimal NLEs on first attempt. Inspired by human learning processes, we introduce Cross-Refine, which employs role modeling by deploying two LLMs as generator and critic, respectively. The generator outputs a first NLE and then refines this initial explanation using feedback and suggestions provided by the critic. Cross-Refine does not require any supervised training data or additional training. We validate Cross-Refine across three NLP tasks using three state-of-the-art open-source LLMs through automatic and human evaluation. We select Self-Refine (Madaan et al., 2023) as the baseline, which only utilizes self-feedback to refine the explanations. Our findings from automatic evaluation and a user study indicate that Cross-Refine outperforms Self-Refine. Meanwhile, Cross-Refine can perform effectively with less powerful LLMs, whereas Self-Refine only yields strong results with ChatGPT. Additionally, we conduct an ablation study to assess the importance of feedback and suggestions. Both of them play an important role in refining explanations. We further evaluate Cross-Refine on a bilingual dataset in English and German.

Given the limited computational and financial resources of news agencies, real-life usage of fact-checking systems requires fast response times. For this reason, our submission to the FEVER-8 claim verification shared task focuses on optimizing the efficiency of such pipelines built around subtasks such as evidence retrieval and veracity prediction. We propose the Semantic Filtering for Efficient Fact Checking (SFEFC) strategy, which is inspired by the FEVER-8 baseline and designed with the goal of reducing the number of LLM calls and other computationally expensive subroutines. Furthermore, we explore the reuse of cosine similarities initially calculated within a dense retrieval step to retrieve the top 10 most relevant evidence sentence sets. We use these sets for semantic filtering methods based on similarity scores and create filters for particularly hard classification labels “Not Enough Information” and “Conflicting Evidence/Cherrypicking” by identifying thresholds for potentially relevant information and the semantic variance within these sets. Compared to the parallelized FEVER-8 baseline, which takes 33.88 seconds on average to process a claim according to the FEVER-8 shared task leaderboard, our non-parallelized system remains competitive in regard to AVeriTeC retrieval scores while reducing the runtime to 7.01 seconds, achieving the fastest average runtime per claim.

Counterfactual examples are widely used in natural language processing (NLP) as valuable data to improve models, and in explainable artificial intelligence (XAI) to understand model behavior. The automated generation of counterfactual examples remains a challenging task even for large language models (LLMs), despite their impressive performance on many tasks. In this paper, we first introduce ZeroCF, a faithful approach for leveraging important words derived from feature attribution methods to generate counterfactual examples in a zero-shot setting. Second, we present a new framework, FitCF, which further verifies aforementioned counterfactuals by label flip verification and then inserts them as demonstrations for few-shot prompting, outperforming three state-of-the-art baselines. Through ablation studies, we identify the importance of each of FitCF’s core components in improving the quality of counterfactuals, as assessed through flip rate, perplexity, and similarity measures. Furthermore, we show the effectiveness of LIME and Integrated Gradients as backbone attribution methods for FitCF and find that the number of demonstrations has the largest effect on performance. Finally, we reveal a strong correlation between the faithfulness of feature attribution scores and the quality of generated counterfactuals, which we hope will serve as an importantfinding for future research in this direction.

Conversational explainable artificial intelligence (ConvXAI) systems based on large language models (LLMs) have garnered considerable attention for their ability to enhance user comprehension through dialogue-based explanations. Current ConvXAI systems often are based on intent recognition to accurately identify the user’s desired intention and map it to an explainability method. While such methods offer great precision and reliability in discerning users’ underlying intentions for English, a significant challenge in the scarcity of training data persists, which impedes multilingual generalization. Besides, the support for free-form custom inputs, which are user-defined data distinct from pre-configured dataset instances, remains largely limited. To bridge these gaps, we first introduce MultiCoXQL, a multilingual extension of the CoXQL dataset spanning five typologically diverse languages, including one low-resource language. Subsequently, we propose a new parsing approach aimed at enhancing multilingual parsing performance, and evaluate three LLMs on MultiCoXQL using various parsing strategies. Furthermore, we present Compass, a new multilingual dataset designed for custom input extraction in ConvXAI systems, encompassing 11 intents across the same five languages as MultiCoXQL. We conduct monolingual, cross-lingual, and multilingual evaluations on Compass, employing three LLMs of varying sizes alongside BERT-type models.

Large Language Models are increasingly used in applications requiring objective assessment, which could be compromised by political bias. Many studies found preferences for left-leaning positions in LLMs, but downstream effects on tasks like fact-checking remain underexplored. In this study, we systematically investigate political bias through exchanging words with euphemisms or dysphemisms in German claims. We construct minimal pairs of factually equivalent claims that differ in political connotation, to assess the consistency of LLMs in classifying them as true or false. We evaluate six LLMs and find that, more than political leaning, the presence of judgmental words significantly influences truthfulness assessment. While a few models show tendencies of political bias, this is not mitigated by explicitly calling for objectivism in prompts. Warning: This paper contains content that may be offensive or upsetting.

Counterfactual examples are widely employed to enhance the performance and robustness of large language models (LLMs) through counterfactual data augmentation (CDA). However, the selection of the judge model used to evaluate label flipping, the primary metric for assessing the validity of generated counterfactuals for CDA, yields inconsistent results. To decipher this, we define four types of relationships between the counterfactual generator and judge models: being the same model, belonging to the same model family, being independent models, and having an distillation relationship. Through extensive experiments involving two state-of-the-art LLM-based methods, three datasets, four generator models, and 15 judge models, complemented by a user study (n = 90), we demonstrate that judge models with an independent, non-fine-tuned relationship to the generator model provide the most reliable label flipping evaluations. Relationships between the generator and judge models, which are closely aligned with the user study for CDA, result in better model performance and robustness. Nevertheless, we find that the gap between the most effective judge models and the results obtained from the user study remains considerably large. This suggests that a fully automated pipeline for CDA may be inadequate and requires human intervention.

Propaganda detection on social media remains challenging due to task complexity and limited high-quality labeled data. This paper introduces a novel framework that combines human expertise with Large Language Model (LLM) assistance to improve both annotation consistency and scalability. We propose a hierarchical taxonomy that organizes 14 fine-grained propaganda techniques (CITATION) into three broader categories, conduct a human annotation study on the HQP dataset (CITATION) that reveals low inter-annotator agreement for fine-grained labels, and implement an LLM-assisted pre-annotation pipeline that extracts propagandistic spans, generates concise explanations, and assigns local labels as well as a global label. A secondary human verification study shows significant improvements in both agreement and time-efficiency. Building on this, we fine-tune smaller language models (SLMs) to perform structured annotation. Instead of fine-tuning on human annotations, we train on high-quality LLM-generated data, allowing a large model to produce these annotations and a smaller model to learn to generate them via knowledge distillation. Our work contributes towards the development of scalable and robust propaganda detection systems, supporting the idea of transparent and accountable media ecosystems in line with SDG 16. The code is publicly available at our GitHub repository.

The overwhelming volume of content being published at any given moment poses a significant challenge for the design of automated fact-checking (AFC) systems on social media, requiring an emphasized consideration of efficiency aspects.As in other fields, systems built upon LLMs have achieved good results on different AFC benchmarks. However, the application of LLMs is accompanied by high resource requirements. The energy consumption of LLMs poses a significant challenge from an ecological perspective, while remaining a bottleneck in latency-sensitive scenarios like AFC within social media. Therefore, we propose a system built upon fine-tuned smaller BERT-based models. When evaluated on the ClimateCheck dataset against decoder-only LLMs, our best fine-tuned model outperforms Phi 4 and approaches Qwen3 14B in reasoning mode — while significantly reducing runtime per claim. Our findings demonstrate that small encoder-only models fine-tuned for specific tasks can still provide a substantive alternative to large decoder-only LLMs, especially in efficiency-concerned settings.

This paper addresses the challenges of sentiment analysis in Ukrainian social media, where users frequently engage in code-switching with Russian and other languages. We introduce COSMUS (COde-Switched MUltilingual Sentiment for Ukrainian Social media), a 12,224-texts corpus collected from Telegram channels, product‐review sites and open datasets, annotated into positive, negative, neutral and mixed sentiment classes as well as language labels (Ukrainian, Russian, code-switched). We benchmark three modeling paradigms: (i) few‐shot prompting of GPT‐4o and DeepSeek V2-chat, (ii) multilingual mBERT, and (iii) the Ukrainian‐centric UkrRoberta. We also analyze calibration and LIME scores of the latter two solutions to verify its performance on various language labels. To mitigate data sparsity we test two augmentation strategies: back‐translation consistently hurts performance, whereas a Large Language Model (LLM) word‐substitution scheme yields up to +2.2% accuracy. Our work delivers the first publicly available dataset and comprehensive benchmark for sentiment classification in Ukrainian code‐switching media. Results demonstrate that language‐specific pre‐training combined with targeted augmentation yields the most accurate and trustworthy predictions in this challenging low‐resource setting.

In an era where political discourse infiltrates online platforms and news media, identifying opinion is increasingly critical, especially in news articles, where objectivity is expected. Readers frequently encounter authors’ inherent political viewpoints, challenging them to discern facts from opinions. Classifying text on a spectrum from left to right is a key task for uncovering these viewpoints. Previous approaches rely on outdated datasets to classify current articles, neglecting that political opinions on certain subjects change over time. This paper explores a novel methodology for detecting political leaning in news articles by augmenting them with political speeches specific to the topic and publication time. We evaluated the impact of the augmentation using BERT and Mistral models. The results show that the BERT model’s F1 score improved from a baseline of 0.82 to 0.85, while the Mistral model’s F1 score increased from 0.30 to 0.31.

With the rise of Large Generative AI Models (LGAIMs), disinformation online has become more concerning than ever before. Within the super-election year 2024, the influence of mis- and disinformation can severely influence public opinion. To combat the increasing amount of disinformation online, humans need to be supported by AI-based tools to increase the effectiveness of detecting false content. This paper examines the critical intersection of the AI Act with the deployment of LGAIMs for disinformation detection and the implications from research, deployer, and the user’s perspective. The utilization of LGAIMs for disinformation detection falls under the high-risk category defined in the AI Act, leading to several obligations that need to be followed after the enforcement of the AI Act. Among others, the obligations include risk management, transparency, and human oversight which pose the challenge of finding adequate technical interpretations. Furthermore, the paper articulates the necessity for clear guidelines and standards that enable the effective, ethical, and legally compliant use of AI. The paper contributes to the discourse on balancing technological advancement with ethical and legal imperatives, advocating for a collaborative approach to utilizing LGAIMs in safeguarding information integrity and fostering trust in digital ecosystems.