Preslav Nakov

Neha Sengupta, Sunil Kumar Sahu, Bokang Jia et al. · berkeley

We introduce Jais and Jais-chat, new state-of-the-art Arabic-centric foundation and instruction-tuned open generative large language models (LLMs). The models are based on the GPT-3 decoder-only architecture and are pretrained on a mixture of Arabic and English texts, including source code in various programming languages. With 13 billion parameters, they demonstrate better knowledge and reasoning capabilities in Arabic than any existing open Arabic and multilingual models by a sizable margin, based on extensive evaluation. Moreover, the models are competitive in English compared to English-centric open models of similar size, despite being trained on much less English data. We provide a detailed description of the training, the tuning, the safety alignment, and the evaluation of the models. We release two open versions of the model -- the foundation Jais model, and an instruction-tuned Jais-chat variant -- with the aim of promoting research on Arabic LLMs. Available at https://huggingface.co/inception-mbzuai/jais-13b-chat

Muhammad Arslan Manzoor, Sarah Albarri, Ziting Xian et al. · berkeley

Multimodality Representation Learning, as a technique of learning to embed information from different modalities and their correlations, has achieved remarkable success on a variety of applications, such as Visual Question Answering (VQA), Natural Language for Visual Reasoning (NLVR), and Vision Language Retrieval (VLR). Among these applications, cross-modal interaction and complementary information from different modalities are crucial for advanced models to perform any multimodal task, e.g., understand, recognize, retrieve, or generate optimally. Researchers have proposed diverse methods to address these tasks. The different variants of transformer-based architectures performed extraordinarily on multiple modalities. This survey presents the comprehensive literature on the evolution and enhancement of deep learning multimodal architectures to deal with textual, visual and audio features for diverse cross-modal and modern multimodal tasks. This study summarizes the (i) recent task-specific deep learning methodologies, (ii) the pretraining types and multimodal pretraining objectives, (iii) from state-of-the-art pretrained multimodal approaches to unifying architectures, and (iv) multimodal task categories and possible future improvements that can be devised for better multimodal learning. Moreover, we prepare a dataset section for new researchers that covers most of the benchmarks for pretraining and finetuning. Finally, major challenges, gaps, and potential research topics are explored. A constantly-updated paperlist related to our survey is maintained at https://github.com/marslanm/multimodality-representation-learning.

Qisheng Liao, Meiting Lai, Preslav Nakov · berkeley

This paper describes our system for SemEval-2023 Task 3 Subtask 2 on Framing Detection. We used a multi-label contrastive loss for fine-tuning large pre-trained language models in a multi-lingual setting, achieving very competitive results: our system was ranked first on the official test set and on the official shared task leaderboard for five of the six languages for which we had training data and for which we could perform fine-tuning. Here, we describe our experimental setup, as well as various ablation studies. The code of our system is available at https://github.com/QishengL/SemEval2023

Yuxia Wang, Haonan Li, Xudong Han et al.

With the rapid evolution of large language models (LLMs), new and hard-to-predict harmful capabilities are emerging. This requires developers to be able to identify risks through the evaluation of "dangerous capabilities" in order to responsibly deploy LLMs. In this work, we collect the first open-source dataset to evaluate safeguards in LLMs, and deploy safer open-source LLMs at a low cost. Our dataset is curated and filtered to consist only of instructions that responsible language models should not follow. We annotate and assess the responses of six popular LLMs to these instructions. Based on our annotation, we proceed to train several BERT-like classifiers, and find that these small classifiers can achieve results that are comparable with GPT-4 on automatic safety evaluation. Warning: this paper contains example data that may be offensive, harmful, or biased.

Zihui Gu, Ju Fan, Nan Tang et al. · berkeley

Fact verification has attracted a lot of research attention recently, e.g., in journalism, marketing, and policymaking, as misinformation and disinformation online can sway one's opinion and affect one's actions. While fact-checking is a hard task in general, in many cases, false statements can be easily debunked based on analytics over tables with reliable information. Hence, table-based fact verification has recently emerged as an important and growing research area. Yet, progress has been limited due to the lack of datasets that can be used to pre-train language models (LMs) to be aware of common table operations, such as aggregating a column or comparing tuples. To bridge this gap, in this paper we introduce PASTA, a novel state-of-the-art framework for table-based fact verification via pre-training with synthesized sentence-table cloze questions. In particular, we design six types of common sentence-table cloze tasks, including Filter, Aggregation, Superlative, Comparative, Ordinal, and Unique, based on which we synthesize a large corpus consisting of 1.2 million sentence-table pairs from WikiTables. PASTA uses a recent pre-trained LM, DeBERTaV3, and further pretrains it on our corpus. Our experimental results show that PASTA achieves new state-of-the-art performance on two table-based fact verification benchmarks: TabFact and SEM-TAB-FACTS. In particular, on the complex set of TabFact, which contains multiple operations, PASTA largely outperforms the previous state of the art by 4.7 points (85.6% vs. 80.9%), and the gap between PASTA and human performance on the small TabFact test set is narrowed to just 1.5 points (90.6% vs. 92.1%).

Shivam Sharma, Siddhant Agarwal, Tharun Suresh et al. · berkeley

Memes are powerful means for effective communication on social media. Their effortless amalgamation of viral visuals and compelling messages can have far-reaching implications with proper marketing. Previous research on memes has primarily focused on characterizing their affective spectrum and detecting whether the meme's message insinuates any intended harm, such as hate, offense, racism, etc. However, memes often use abstraction, which can be elusive. Here, we introduce a novel task - EXCLAIM, generating explanations for visual semantic role labeling in memes. To this end, we curate ExHVV, a novel dataset that offers natural language explanations of connotative roles for three types of entities - heroes, villains, and victims, encompassing 4,680 entities present in 3K memes. We also benchmark ExHVV with several strong unimodal and multimodal baselines. Moreover, we posit LUMEN, a novel multimodal, multi-task learning framework that endeavors to address EXCLAIM optimally by jointly learning to predict the correct semantic roles and correspondingly to generate suitable natural language explanations. LUMEN distinctly outperforms the best baseline across 18 standard natural language generation evaluation metrics. Our systematic evaluation and analyses demonstrate that characteristic multimodal cues required for adjudicating semantic roles are also helpful for generating suitable explanations.

Yuxia Wang, Revanth Gangi Reddy, Zain Muhammad Mujahid et al.

The increased use of large language models (LLMs) across a variety of real-world applications calls for mechanisms to verify the factual accuracy of their outputs. In this work, we present a holistic end-to-end solution for annotating the factuality of LLM-generated responses, which encompasses a multi-stage annotation scheme designed to yield detailed labels concerning the verifiability and factual inconsistencies found in LLM outputs. We further construct an open-domain document-level factuality benchmark in three-level granularity: claim, sentence and document, aiming to facilitate the evaluation of automatic fact-checking systems. Preliminary experiments show that FacTool, FactScore and Perplexity.ai are struggling to identify false claims, with the best F1=0.63 by this annotation solution based on GPT-4. Annotation tool, benchmark and code are available at https://github.com/yuxiaw/Factcheck-GPT.

Momchil Hardalov, Pepa Atanasova, Todor Mihaylov et al. · berkeley

We present bgGLUE(Bulgarian General Language Understanding Evaluation), a benchmark for evaluating language models on Natural Language Understanding (NLU) tasks in Bulgarian. Our benchmark includes NLU tasks targeting a variety of NLP problems (e.g., natural language inference, fact-checking, named entity recognition, sentiment analysis, question answering, etc.) and machine learning tasks (sequence labeling, document-level classification, and regression). We run the first systematic evaluation of pre-trained language models for Bulgarian, comparing and contrasting results across the nine tasks in the benchmark. The evaluation results show strong performance on sequence labeling tasks, but there is a lot of room for improvement for tasks that require more complex reasoning. We make bgGLUE publicly available together with the fine-tuning and the evaluation code, as well as a public leaderboard at https://bgglue.github.io/, and we hope that it will enable further advancements in developing NLU models for Bulgarian.

Mervat Abassy, Kareem Elozeiri, Alexander Aziz et al.

The ease of access to large language models (LLMs) has enabled a widespread of machine-generated texts, and now it is often hard to tell whether a piece of text was human-written or machine-generated. This raises concerns about potential misuse, particularly within educational and academic domains. Thus, it is important to develop practical systems that can automate the process. Here, we present one such system, LLM-DetectAIve, designed for fine-grained detection. Unlike most previous work on machine-generated text detection, which focused on binary classification, LLM-DetectAIve supports four categories: (i) human-written, (ii) machine-generated, (iii) machine-written, then machine-humanized, and (iv) human-written, then machine-polished. Category (iii) aims to detect attempts to obfuscate the fact that a text was machine-generated, while category (iv) looks for cases where the LLM was used to polish a human-written text, which is typically acceptable in academic writing, but not in education. Our experiments show that LLM-DetectAIve can effectively identify the above four categories, which makes it a potentially useful tool in education, academia, and other domains. LLM-DetectAIve is publicly accessible at https://github.com/mbzuai-nlp/LLM-DetectAIve. The video describing our system is available at https://youtu.be/E8eT_bE7k8c.

Hasan Iqbal, Yuxia Wang, Minghan Wang et al.

The increased use of large language models (LLMs) across a variety of real-world applications calls for automatic tools to check the factual accuracy of their outputs, as LLMs often hallucinate. This is difficult as it requires assessing the factuality of free-form open-domain responses. While there has been a lot of research on this topic, different papers use different evaluation benchmarks and measures, which makes them hard to compare and hampers future progress. To mitigate these issues, we developed OpenFactCheck, a unified framework, with three modules: (i) RESPONSEEVAL, which allows users to easily customize an automatic fact-checking system and to assess the factuality of all claims in an input document using that system, (ii) LLMEVAL, which assesses the overall factuality of an LLM, and (iii) CHECKEREVAL, a module to evaluate automatic fact-checking systems. OpenFactCheck is open-sourced (https://github.com/mbzuai-nlp/openfactcheck) and publicly released as a Python library (https://pypi.org/project/openfactcheck/) and also as a web service (http://app.openfactcheck.com). A video describing the system is available at https://youtu.be/-i9VKL0HleI.

Yixi Zhou, Fan Zhang, Yu Chen et al.

Financial question answering (QA) over long corporate filings requires evidence to satisfy strict constraints on entities, financial metrics, fiscal periods, and numeric values. However, existing LLM-based rerankers primarily optimize semantic relevance, leading to unstable rankings and opaque decisions on long documents. We propose FinCards, a structured reranking framework that reframes financial evidence selection as constraint satisfaction under a finance-aware schema. FinCards represents filing chunks and questions using aligned schema fields (entities, metrics, periods, and numeric spans), enabling deterministic field-level matching. Evidence is selected via a multi-stage tournament reranking with stability-aware aggregation, producing auditable decision traces. Across two corporate filing QA benchmarks, FinCards substantially improves early-rank retrieval over both lexical and LLM-based reranking baselines, while reducing ranking variance, without requiring model fine-tuning or unpredictable inference budgets. Our code is available at https://github.com/XanderZhou2022/FINCARDS.

Shivam Sharma, Atharva Kulkarni, Tharun Suresh et al. · cmu

Memes can sway people's opinions over social media as they combine visual and textual information in an easy-to-consume manner. Since memes instantly turn viral, it becomes crucial to infer their intent and potentially associated harmfulness to take timely measures as needed. A common problem associated with meme comprehension lies in detecting the entities referenced and characterizing the role of each of these entities. Here, we aim to understand whether the meme glorifies, vilifies, or victimizes each entity it refers to. To this end, we address the task of role identification of entities in harmful memes, i.e., detecting who is the 'hero', the 'villain', and the 'victim' in the meme, if any. We utilize HVVMemes - a memes dataset on US Politics and Covid-19 memes, released recently as part of the CONSTRAINT@ACL-2022 shared-task. It contains memes, entities referenced, and their associated roles: hero, villain, victim, and other. We further design VECTOR (Visual-semantic role dEteCToR), a robust multi-modal framework for the task, which integrates entity-based contextual information in the multi-modal representation and compare it to several standard unimodal (text-only or image-only) or multi-modal (image+text) models. Our experimental results show that our proposed model achieves an improvement of 4% over the best baseline and 1% over the best competing stand-alone submission from the shared-task. Besides divulging an extensive experimental setup with comparative analyses, we finally highlight the challenges encountered in addressing the complex task of semantic role labeling within memes.

Ali Ezzat Shahroor, Mohamed Bayan Kmainasi, Abul Hasnat et al.

Memes are a dominant medium for online communication and manipulation because meaning emerges from interactions between embedded text, imagery, and cultural context. Existing meme research is distributed across tasks (hate, misogyny, propaganda, sentiment, humour) and languages, which limits cross-domain generalization. To address this gap we propose MemeLens, a unified multilingual and multitask explanation-enhanced Vision Language Model (VLM) for meme understanding. We consolidate $38$ public meme datasets, filter and map dataset-specific labels into a shared taxonomy of $20$ tasks spanning harm, targets, figurative/pragmatic intent, and affect. We present a comprehensive empirical analysis across modeling paradigms, task categories, and datasets. Our findings suggest that robust meme understanding requires multimodal training, exhibits substantial variation across semantic categories, and remains sensitive to over-specialization when models are fine-tuned on individual datasets rather than trained in a unified setting. We make the experimental resources (https://github.com/MohamedBayan/MemeLens), model (https://huggingface.co/QCRI/MemeLens-VLM) and datasets (https://huggingface.co/datasets/QCRI/MemeLens) publicly available to the community.

Georgi Pachov, Dimitar Dimitrov, Ivan Koychev et al. · berkeley

The wide-spread use of social networks has given rise to subjective, misleading, and even false information on the Internet. Thus, subjectivity detection can play an important role in ensuring the objectiveness and the quality of a piece of information. This paper presents the solution built by the Gpachov team for the CLEF-2023 CheckThat! lab Task~2 on subjectivity detection. Three different research directions are explored. The first one is based on fine-tuning a sentence embeddings encoder model and dimensionality reduction. The second one explores a sample-efficient few-shot learning model. The third one evaluates fine-tuning a multilingual transformer on an altered dataset, using data from multiple languages. Finally, the three approaches are combined in a simple majority voting ensemble, resulting in 0.77 macro F1 on the test set and achieving 2nd place on the English subtask.

Shubham Mittal, Preslav Nakov · berkeley

We present our system for the two subtasks of the shared task on propaganda detection in Arabic, part of WANLP'2022. Subtask 1 is a multi-label classification problem to find the propaganda techniques used in a given tweet. Our system for this task uses XLM-R to predict probabilities for the target tweet to use each of the techniques. In addition to finding the techniques, Subtask 2 further asks to identify the textual span for each instance of each technique that is present in the tweet; the task can be modeled as a sequence tagging problem. We use a multi-granularity network with mBERT encoder for Subtask 2. Overall, our system ranks second for both subtasks (out of 14 and 3 participants, respectively). Our empirical analysis show that it does not help to use a much larger English corpus annotated with propaganda techniques, regardless of whether used in English or after translation to Arabic.

Tarek Mahmoud, Veronika Solopova, Premtim Sahitaj et al.

Temporal language does more than place events on a timeline. In news discourse, references to the past, present, and future can function as rhetorical devices that shape interpretation and persuasion. Here, we study temporal framing, defined as the persuasive use of time-related language to structure meaning rather than to report chronology. We propose a taxonomy of eight temporal frames grounded in prior work on temporality and framing, and we realize it through expert annotation of a multilingual news corpus. The resulting dataset includes 458 English and German news articles, with over 2K temporally framed sentences and approximately 3K temporal framing annotations identified from a corpus of more than 20K sentences. We analyze frame prevalence, co-occurrence patterns, and lexical cues, and evaluate temporal framing detection using supervised fine-tuning and zero-shot classification. Our experiments show that temporal framing is learnable at the sentence level, with supervised models substantially outperforming zero-shot approaches. We publicly release the corpus to support future research on temporal framing: https://mbzuai-nlp.github.io/temporal-framing/.

Momchil Hardalov, Anton Chernyavskiy, Ivan Koychev et al.

While there has been substantial progress in developing systems to automate fact-checking, they still lack credibility in the eyes of the users. Thus, an interesting approach has emerged: to perform automatic fact-checking by verifying whether an input claim has been previously fact-checked by professional fact-checkers and to return back an article that explains their decision. This is a sensible approach as people trust manual fact-checking, and as many claims are repeated multiple times. Yet, a major issue when building such systems is the small number of known tweet--verifying article pairs available for training. Here, we aim to bridge this gap by making use of crowd fact-checking, i.e., mining claims in social media for which users have responded with a link to a fact-checking article. In particular, we mine a large-scale collection of 330,000 tweets paired with a corresponding fact-checking article. We further propose an end-to-end framework to learn from this noisy data based on modified self-adaptive training, in a distant supervision scenario. Our experiments on the CLEF'21 CheckThat! test set show improvements over the state of the art by two points absolute. Our code and datasets are available at https://github.com/mhardalov/crowdchecked-claims

Xinyuan Lu, Liangming Pan, Yubo Ma et al.

Current Large Language Models (LLMs) exhibit limited ability to understand table structures and to apply precise numerical reasoning, which is crucial for tasks such as table question answering (TQA) and table-based fact verification (TFV). To address these challenges, we introduce our Tool-Augmented Reasoning framework for Tables (TART), which integrates LLMs with specialized tools. TART contains three key components: a table formatter to ensure accurate data representation, a tool maker to develop specific computational tools, and an explanation generator to maintain explainability. We also present the TOOLTAB dataset, a new benchmark designed specifically for training LLMs in table-tool integration. Our experiments indicate that TART achieves substantial improvements over existing methods (e.g., Chain-of-Thought) by improving both the precision of data processing and the clarity of the reasoning process. Notably, TART paired with CodeLlama achieves 90.0% of the accuracy of the closed-sourced LLM GPT-3.5-turbo, highlighting its robustness in diverse real-world scenarios. All the code and data are available at https://github.com/XinyuanLu00/TART.

Isabelle Augenstein, Timothy Baldwin, Meeyoung Cha et al.

The emergence of tools based on Large Language Models (LLMs), such as OpenAI's ChatGPT, Microsoft's Bing Chat, and Google's Bard, has garnered immense public attention. These incredibly useful, natural-sounding tools mark significant advances in natural language generation, yet they exhibit a propensity to generate false, erroneous, or misleading content -- commonly referred to as "hallucinations." Moreover, LLMs can be exploited for malicious applications, such as generating false but credible-sounding content and profiles at scale. This poses a significant challenge to society in terms of the potential deception of users and the increasing dissemination of inaccurate information. In light of these risks, we explore the kinds of technological innovations, regulatory reforms, and AI literacy initiatives needed from fact-checkers, news organizations, and the broader research and policy communities. By identifying the risks, the imminent threats, and some viable solutions, we seek to shed light on navigating various aspects of veracity in the era of generative AI.

Angel Beshirov, Milena Dobreva, Dimitar Dimitrov et al.

The digitization of historical documents is crucial for preserving the cultural heritage of the society. An important step in this process is converting scanned images to text using Optical Character Recognition (OCR), which can enable further search, information extraction, etc. Unfortunately, this is a hard problem as standard OCR tools are not tailored to deal with historical orthography as well as with challenging layouts. Thus, it is standard to apply an additional text correction step on the OCR output when dealing with such documents. In this work, we focus on Bulgarian, and we create the first benchmark dataset for evaluating the OCR text correction for historical Bulgarian documents written in the first standardized Bulgarian orthography: the Drinov orthography from the 19th century. We further develop a method for automatically generating synthetic data in this orthography, as well as in the subsequent Ivanchev orthography, by leveraging vast amounts of contemporary literature Bulgarian texts. We then use state-of-the-art LLMs and encoder-decoder framework which we augment with diagonal attention loss and copy and coverage mechanisms to improve the post-OCR text correction. The proposed method reduces the errors introduced during recognition and improves the quality of the documents by 25\%, which is an increase of 16\% compared to the state-of-the-art on the ICDAR 2019 Bulgarian dataset. We release our data and code at \url{https://github.com/angelbeshirov/post-ocr-text-correction}.}

Rabindra Nath Nandi, Firoj Alam, Preslav Nakov

Harmful or abusive online content has been increasing over time, raising concerns for social media platforms, government agencies, and policymakers. Such harmful or abusive content can have major negative impact on society, e.g., cyberbullying can lead to suicides, rumors about COVID-19 can cause vaccine hesitance, promotion of fake cures for COVID-19 can cause health harms and deaths. The content that is posted and shared online can be textual, visual, or a combination of both, e.g., in a meme. Here, we describe our experiments in detecting the roles of the entities (hero, villain, victim) in harmful memes, which is part of the CONSTRAINT-2022 shared task, as well as our system for the task. We further provide a comparative analysis of different experimental settings (i.e., unimodal, multimodal, attention, and augmentation). For reproducibility, we make our experimental code publicly available. \url{https://github.com/robi56/harmful_memes_block_fusion}

Shivam Sharma, Firoj Alam, Md. Shad Akhtar et al.

The automatic identification of harmful content online is of major concern for social media platforms, policymakers, and society. Researchers have studied textual, visual, and audio content, but typically in isolation. Yet, harmful content often combines multiple modalities, as in the case of memes, which are of particular interest due to their viral nature. With this in mind, here we offer a comprehensive survey with a focus on harmful memes. Based on a systematic analysis of recent literature, we first propose a new typology of harmful memes, and then we highlight and summarize the relevant state of the art. One interesting finding is that many types of harmful memes are not really studied, e.g., such featuring self-harm and extremism, partly due to the lack of suitable datasets. We further find that existing datasets mostly capture multi-class scenarios, which are not inclusive of the affective spectrum that memes can represent. Another observation is that memes can propagate globally through repackaging in different languages and that they can also be multilingual, blending different cultures. We conclude by highlighting several challenges related to multimodal semiotics, technological constraints, and non-trivial social engagement, and we present several open-ended aspects such as delineating online harm and empirically examining related frameworks and assistive interventions, which we believe will motivate and drive future research.

Liangming Pan, Xinyuan Lu, Min-Yen Kan et al. · pku

Fact-checking real-world claims often requires complex, multi-step reasoning due to the absence of direct evidence to support or refute them. However, existing fact-checking systems often lack transparency in their decision-making, making it challenging for users to comprehend their reasoning process. To address this, we propose the Question-guided Multi-hop Fact-Checking (QACHECK) system, which guides the model's reasoning process by asking a series of questions critical for verifying a claim. QACHECK has five key modules: a claim verifier, a question generator, a question-answering module, a QA validator, and a reasoner. Users can input a claim into QACHECK, which then predicts its veracity and provides a comprehensive report detailing its reasoning process, guided by a sequence of (question, answer) pairs. QACHECK also provides the source of evidence supporting each question, fostering a transparent, explainable, and user-friendly fact-checking process. A recorded video of QACHECK is at https://www.youtube.com/watch?v=ju8kxSldM64

Kung-Hsiang Huang, Kathleen McKeown, Preslav Nakov et al.

Despite recent advances in detecting fake news generated by neural models, their results are not readily applicable to effective detection of human-written disinformation. What limits the successful transfer between them is the sizable gap between machine-generated fake news and human-authored ones, including the notable differences in terms of style and underlying intent. With this in mind, we propose a novel framework for generating training examples that are informed by the known styles and strategies of human-authored propaganda. Specifically, we perform self-critical sequence training guided by natural language inference to ensure the validity of the generated articles, while also incorporating propaganda techniques, such as appeal to authority and loaded language. In particular, we create a new training dataset, PropaNews, with 2,256 examples, which we release for future use. Our experimental results show that fake news detectors trained on PropaNews are better at detecting human-written disinformation by 3.62 - 7.69% F1 score on two public datasets.

Abdelaziz Bounhar, Rania Hossam Elmohamady Elbadry, Hadi Abdine et al.

Steering Large Language Models (LLMs) through activation interventions has emerged as a lightweight alternative to fine-tuning for alignment and personalization. Recent work on Bi-directional Preference Optimization (BiPO) shows that dense steering vectors can be learned directly from preference data in a Direct Preference Optimization (DPO) fashion, enabling control over truthfulness, hallucinations, and safety behaviors. However, dense steering vectors often entangle multiple latent factors due to neuron multi-semanticity, limiting their effectiveness and stability in fine-grained settings such as cultural alignment, where closely related values and behaviors (e.g., among Middle Eastern cultures) must be distinguished. In this paper, we propose Yet another Policy Optimization (YaPO), a \textit{reference-free} method that learns \textit{sparse steering vectors} in the latent space of a Sparse Autoencoder (SAE). By optimizing sparse codes, YaPO produces disentangled, interpretable, and efficient steering directions. Empirically, we show that YaPO converges faster, achieves stronger performance, and exhibits improved training stability compared to dense steering baselines. Beyond cultural alignment, YaPO generalizes to a range of alignment-related behaviors, including hallucination, wealth-seeking, jailbreak, and power-seeking. Importantly, YaPO preserves general knowledge, with no measurable degradation on MMLU. Overall, our results show that YaPO provides a general recipe for efficient, stable, and fine-grained alignment of LLMs, with broad applications to controllability and domain adaptation. The associated code and data are publicly available\footnote{https://github.com/MBZUAI-Paris/YaPO}.

Yixi Zhou, Fan Zhang, Zhiqiao Guo et al.

Despite strong performance on Text-to-SQL benchmarks, it remains unclear whether LLM-generated SQL programs are structurally reliable. In this work, we investigate the structural behavior of LLM-generated SQL queries and introduce SQLStructEval, a framework for analyzing program structures through canonical abstract syntax tree (AST) representations. Our experiments on the Spider benchmark show that modern LLMs often produce structurally diverse queries for the same input, even when execution results are correct, and that such variance is frequently triggered by surface-level input changes such as paraphrases or schema presentation. We further show that generating queries in a structured space via a compile-style pipeline can improve both execution accuracy and structural consistency. These findings suggest that structural reliability is a critical yet overlooked dimension for evaluating LLM-based program generation systems. Our code is available at https://anonymous.4open.science/r/StructEval-2435.

Sanjay Chawla, Preslav Nakov, Ahmed Ali et al. · berkeley

It is ten years since neural networks made their spectacular comeback. Prompted by this anniversary, we take a holistic perspective on Artificial Intelligence (AI). Supervised Learning for cognitive tasks is effectively solved - provided we have enough high-quality labeled data. However, deep neural network models are not easily interpretable, and thus the debate between blackbox and whitebox modeling has come to the fore. The rise of attention networks, self-supervised learning, generative modeling, and graph neural networks has widened the application space of AI. Deep Learning has also propelled the return of reinforcement learning as a core building block of autonomous decision making systems. The possible harms made possible by new AI technologies have raised socio-technical issues such as transparency, fairness, and accountability. The dominance of AI by Big-Tech who control talent, computing resources, and most importantly, data may lead to an extreme AI divide. Failure to meet high expectations in high profile, and much heralded flagship projects like self-driving vehicles could trigger another AI winter.

Jiahui Geng, Fengyu Cai, Yuxia Wang et al.

Large language models (LLMs) have demonstrated remarkable capabilities across a wide range of tasks in various domains. Despite their impressive performance, they can be unreliable due to factual errors in their generations. Assessing their confidence and calibrating them across different tasks can help mitigate risks and enable LLMs to produce better generations. There has been a lot of recent research aiming to address this, but there has been no comprehensive overview to organize it and outline the main lessons learned. The present survey aims to bridge this gap. In particular, we outline the challenges and we summarize recent technical advancements for LLM confidence estimation and calibration. We further discuss their applications and suggest promising directions for future work.

Serena Tardelli, Leonardo Nizzoli, Maurizio Tesconi et al.

Large-scale online campaigns, malicious or otherwise, require a significant degree of coordination among participants, which sparked interest in the study of coordinated online behavior. State-of-the-art methods for detecting coordinated behavior perform static analyses, disregarding the temporal dynamics of coordination. Here, we carry out the first dynamic analysis of coordinated behavior. To reach our goal we build a multiplex temporal network and we perform dynamic community detection to identify groups of users that exhibited coordinated behaviors in time. Thanks to our novel approach we find that: (i) coordinated communities feature variable degrees of temporal instability; (ii) dynamic analyses are needed to account for such instability, and results of static analyses can be unreliable and scarcely representative of unstable communities; (iii) some users exhibit distinct archetypal behaviors that have important practical implications; (iv) content and network characteristics contribute to explaining why users leave and join coordinated communities. Our results demonstrate the advantages of dynamic analyses and open up new directions of research on the unfolding of online debates, on the strategies of coordinated communities, and on the patterns of online influence.

Daniil Orel, Dilshod Azizov, Indraneil Paul et al.

Large language models (LLMs) are increasingly capable of generating functional source code, raising concerns about authorship, accountability, and security. While detecting AI-generated code is critical, existing datasets and benchmarks are narrow, typically limited to binary human-machine classification under in-distribution settings. To bridge this gap, we introduce $\emph{AICD Bench}$, the most comprehensive benchmark for AI-generated code detection. It spans $\emph{2M examples}$, $\emph{77 models}$ across $\emph{11 families}$, and $\emph{9 programming languages}$, including recent reasoning models. Beyond scale, AICD Bench introduces three realistic detection tasks: ($\emph{i}$)~$\emph{Robust Binary Classification}$ under distribution shifts in language and domain, ($\emph{ii}$)~$\emph{Model Family Attribution}$, grouping generators by architectural lineage, and ($\emph{iii}$)~$\emph{Fine-Grained Human-Machine Classification}$ across human, machine, hybrid, and adversarial code. Extensive evaluation on neural and classical detectors shows that performance remains far below practical usability, particularly under distribution shift and for hybrid or adversarial code. We release AICD Bench as a $\emph{unified, challenging evaluation suite}$ to drive the next generation of robust approaches for AI-generated code detection. The data and the code are available at https://huggingface.co/AICD-bench}.

Zihang Fu, Fanxiao Li, Jianyang Gu et al.

Large Language Model (LLM)-augmented Community Notes offer a scalable path for timely, evidence-grounded correction of health misinformation on social platforms. However, they still reset at every post, leaving useful correction experience from prior cases unused. We introduce EvoNote, an agentic framework that enables health Community Notes generation to self-evolve through an evolving experience memory of prior misinformation correction episodes. Its core is fine-grained credit assignment: EvoNote grounds trajectory-level feedback in health-specific note qualities and distills it into action-level memory for claim analysis, evidence acquisition, and note writing. We evaluate EvoNote on MM-HealthCN, a 1.2K-instance multimodal benchmark of user-flagged health posts with human-written Community Notes and crowd-derived helpfulness labels. Under a human-validated hierarchical utility judge, EvoNote-generated notes are preferred over corresponding human-written notes in 89.6% of cases; on a separate set of Needs More Ratings posts without a crowd helpfulness verdict, EvoNote produces helpful notes for 82.0% of cases. It also reduces the median time needed to produce a candidate correction from over 13 hours in the human-note pipeline to under 2 minutes. Analyses link these gains to stronger evidence use and reusable correction strategies, positioning self-evolving note generation as a promising paradigm for health misinformation governance.

Firoj Alam, Hamdy Mubarak, Wajdi Zaghouani et al.

Propaganda is the expression of an opinion or an action by an individual or a group deliberately designed to influence the opinions or the actions of other individuals or groups with reference to predetermined ends, which is achieved by means of well-defined rhetorical and psychological devices. Propaganda techniques are commonly used in social media to manipulate or to mislead users. Thus, there has been a lot of recent research on automatic detection of propaganda techniques in text as well as in memes. However, so far the focus has been primarily on English. With the aim to bridge this language gap, we ran a shared task on detecting propaganda techniques in Arabic tweets as part of the WANLP 2022 workshop, which included two subtasks. Subtask~1 asks to identify the set of propaganda techniques used in a tweet, which is a multilabel classification problem, while Subtask~2 asks to detect the propaganda techniques used in a tweet together with the exact span(s) of text in which each propaganda technique appears. The task attracted 63 team registrations, and eventually 14 and 3 teams made submissions for subtask 1 and 2, respectively. Finally, 11 teams submitted system description papers.

Maram Hasanain, Firoj Alam, Hamdy Mubarak et al.

We present an overview of the ArAIEval shared task, organized as part of the first ArabicNLP 2023 conference co-located with EMNLP 2023. ArAIEval offers two tasks over Arabic text: (i) persuasion technique detection, focusing on identifying persuasion techniques in tweets and news articles, and (ii) disinformation detection in binary and multiclass setups over tweets. A total of 20 teams participated in the final evaluation phase, with 14 and 16 teams participating in Tasks 1 and 2, respectively. Across both tasks, we observed that fine-tuning transformer models such as AraBERT was at the core of the majority of the participating systems. We provide a description of the task setup, including a description of the dataset construction and the evaluation setup. We further give a brief overview of the participating systems. All datasets and evaluation scripts from the shared task are released to the research community. (https://araieval.gitlab.io/) We hope this will enable further research on these important tasks in Arabic.

Shivam Sharma, Md. Shad Akhtar, Preslav Nakov et al.

Internet memes have emerged as an increasingly popular means of communication on the Web. Although typically intended to elicit humour, they have been increasingly used to spread hatred, trolling, and cyberbullying, as well as to target specific individuals, communities, or society on political, socio-cultural, and psychological grounds. While previous work has focused on detecting harmful, hateful, and offensive memes, identifying whom they attack remains a challenging and underexplored area. Here we aim to bridge this gap. In particular, we create a dataset where we annotate each meme with its victim(s) such as the name of the targeted person(s), organization(s), and community(ies). We then propose DISARM (Detecting vIctimS targeted by hARmful Memes), a framework that uses named entity recognition and person identification to detect all entities a meme is referring to, and then, incorporates a novel contextualized multimodal deep neural network to classify whether the meme intends to harm these entities. We perform several systematic experiments on three test setups, corresponding to entities that are (a) all seen while training, (b) not seen as a harmful target on training, and (c) not seen at all on training. The evaluation results show that DISARM significantly outperforms ten unimodal and multimodal systems. Finally, we show that DISARM is interpretable and comparatively more generalizable and that it can reduce the relative error rate for harmful target identification by up to 9 points absolute over several strong multimodal rivals.

Elitsa Yotkova, Violeta Kastreva, Dimitar Dimitrov et al.

SemEval-2026 Task 13 investigates machine-generated code detection across multiple programming languages and application scenarios, asking participating systems to generalize to unseen languages and domains. This paper describes our participation in Subtask A (binary classification) and explores both pretrained code encoders and lightweight feature-based methods. We design ratio-based features that are less sensitive to snippet length. To support the extraction of descriptiveness-related signals, we use parsing engines and a programming-language classifier. Additionally, we train a separate code-vs-text line classifier to identify raw natural language segments embedded within samples. We combine a shallow decision tree with heuristic rules derived from data analysis to produce the final predictions. Our approach is computationally efficient, requires only CPU resources for training, and achieves near-instant inference time, offering a lightweight alternative to large pretrained models.

Jinyan Su, Claire Cardie, Preslav Nakov

In the age of large language models (LLMs) and the widespread adoption of AI-driven content creation, the landscape of information dissemination has witnessed a paradigm shift. With the proliferation of both human-written and machine-generated real and fake news, robustly and effectively discerning the veracity of news articles has become an intricate challenge. While substantial research has been dedicated to fake news detection, this either assumes that all news articles are human-written or abruptly assumes that all machine-generated news are fake. Thus, a significant gap exists in understanding the interplay between machine-(paraphrased) real news, machine-generated fake news, human-written fake news, and human-written real news. In this paper, we study this gap by conducting a comprehensive evaluation of fake news detectors trained in various scenarios. Our primary objectives revolve around the following pivotal question: How to adapt fake news detectors to the era of LLMs? Our experiments reveal an interesting pattern that detectors trained exclusively on human-written articles can indeed perform well at detecting machine-generated fake news, but not vice versa. Moreover, due to the bias of detectors against machine-generated texts \cite{su2023fake}, they should be trained on datasets with a lower machine-generated news ratio than the test set. Building on our findings, we provide a practical strategy for the development of robust fake news detectors.

Megha Sundriyal, Tanmoy Chakraborty, Preslav Nakov

With the rise of social media, users are exposed to many misleading claims. However, the pervasive noise inherent in these posts presents a challenge in identifying precise and prominent claims that require verification. Extracting the important claims from such posts is arduous and time-consuming, yet it is an underexplored problem. Here, we aim to bridge this gap. We introduce a novel task, Claim Normalization (aka ClaimNorm), which aims to decompose complex and noisy social media posts into more straightforward and understandable forms, termed normalized claims. We propose CACN, a pioneering approach that leverages chain-of-thought and claim check-worthiness estimation, mimicking human reasoning processes, to comprehend intricate claims. Moreover, we capitalize on the in-context learning capabilities of large language models to provide guidance and to improve claim normalization. To evaluate the effectiveness of our proposed model, we meticulously compile a comprehensive real-world dataset, CLAN, comprising more than 6k instances of social media posts alongside their respective normalized claims. Our experiments demonstrate that CACN outperforms several baselines across various evaluation measures. Finally, our rigorous error analysis validates CACN's capabilities and pitfalls.

Guokan Shang, Hadi Abdine, Yousef Khoubrane et al.

We introduce Atlas-Chat, the first-ever collection of LLMs specifically developed for dialectal Arabic. Focusing on Moroccan Arabic, also known as Darija, we construct our instruction dataset by consolidating existing Darija language resources, creating novel datasets both manually and synthetically, and translating English instructions with stringent quality control. Atlas-Chat-2B, 9B, and 27B models, fine-tuned on the dataset, exhibit superior ability in following Darija instructions and performing standard NLP tasks. Notably, our models outperform both state-of-the-art and Arabic-specialized LLMs like LLaMa, Jais, and AceGPT, e.g., our 9B model gains a 13% performance boost over a larger 13B model on DarijaMMLU, in our newly introduced evaluation suite for Darija covering both discriminative and generative tasks. Furthermore, we perform an experimental analysis of various fine-tuning strategies and base model choices to determine optimal configurations. All our resources are publicly accessible, and we believe our work offers comprehensive design methodologies of instruction-tuning for low-resource languages, which are often neglected in favor of data-rich languages by contemporary LLMs.

Jinyan Su, Terry Yue Zhuo, Jonibek Mansurov et al.

The spread of fake news has emerged as a critical challenge, undermining trust and posing threats to society. In the era of Large Language Models (LLMs), the capability to generate believable fake content has intensified these concerns. In this study, we present a novel paradigm to evaluate fake news detectors in scenarios involving both human-written and LLM-generated misinformation. Intriguingly, our findings reveal a significant bias in many existing detectors: they are more prone to flagging LLM-generated content as fake news while often misclassifying human-written fake news as genuine. This unexpected bias appears to arise from distinct linguistic patterns inherent to LLM outputs. To address this, we introduce a mitigation strategy that leverages adversarial training with LLM-paraphrased genuine news. The resulting model yielded marked improvements in detection accuracy for both human and LLM-generated news. To further catalyze research in this domain, we release two comprehensive datasets, \texttt{GossipCop++} and \texttt{PolitiFact++}, thus amalgamating human-validated articles with LLM-generated fake and real news.

Muhammad Arslan Manzoor, Dilshod Azizov, Daniil Orel et al.

News outlets shape public opinion at a scale that makes automated detection of political bias and factuality essential. However, the field still lacks unified resources, comprehensive evaluations across diverse approaches, and systematic analyses of the representations and fusion strategies that matter most, especially under label sparsity and dataset diversity. In addition, there is little empirical work reporting broad, observation-driven findings about what consistently works, what fails, and why. We address these gaps through four main contributions. First, we introduce MBFC-2025, a large-scale label set covering approximately 2,600 outlets from Media Bias/Fact Check (MBFC). Second, we construct multiview representations for ACL-2020 (Panayotov et al., 2022), which includes around 900 outlets, as well as for MBFC-2025. These representations span Alexa graphs, hyperlink graphs, LLM-derived graphs, articles, and Wikipedia descriptions. Third, we provide a systematic evaluation and analysis of embedding views and fusion strategies, including a reinforcement learning-based fusion variant. Fourth, we conduct extensive experiments that achieve state-of-the-art results on ACL-2020 and establish strong benchmarks on MBFC-2025.

Panayot Panayotov, Utsav Shukla, Husrev Taha Sencar et al.

We study the problem of profiling news media on the Web with respect to their factuality of reporting and bias. This is an important but under-studied problem related to disinformation and "fake news" detection, but it addresses the issue at a coarser granularity compared to looking at an individual article or an individual claim. This is useful as it allows to profile entire media outlets in advance. Unlike previous work, which has focused primarily on text (e.g.,~on the text of the articles published by the target website, or on the textual description in their social media profiles or in Wikipedia), here our main focus is on modeling the similarity between media outlets based on the overlap of their audience. This is motivated by homophily considerations, i.e.,~the tendency of people to have connections to people with similar interests, which we extend to media, hypothesizing that similar types of media would be read by similar kinds of users. In particular, we propose GREENER (GRaph nEural nEtwork for News mEdia pRofiling), a model that builds a graph of inter-media connections based on their audience overlap, and then uses graph neural networks to represent each medium. We find that such representations are quite useful for predicting the factuality and the bias of news media outlets, yielding improvements over state-of-the-art results reported on two datasets. When augmented with conventionally used representations obtained from news articles, Twitter, YouTube, Facebook, and Wikipedia, prediction accuracy is found to improve by 2.5-27 macro-F1 points for the two tasks.

Shubham Mittal, Megha Sundriyal, Preslav Nakov

Claim span identification (CSI) is an important step in fact-checking pipelines, aiming to identify text segments that contain a checkworthy claim or assertion in a social media post. Despite its importance to journalists and human fact-checkers, it remains a severely understudied problem, and the scarce research on this topic so far has only focused on English. Here we aim to bridge this gap by creating a novel dataset, X-CLAIM, consisting of 7K real-world claims collected from numerous social media platforms in five Indian languages and English. We report strong baselines with state-of-the-art encoder-only language models (e.g., XLM-R) and we demonstrate the benefits of training on multiple languages over alternative cross-lingual transfer methods such as zero-shot transfer, or training on translated data, from a high-resource language such as English. We evaluate generative large language models from the GPT series using prompting methods on the X-CLAIM dataset and we find that they underperform the smaller encoder-only language models for low-resource languages.

Jiateng Liu, Lin Ai, Zizhou Liu et al.

Propaganda plays a critical role in shaping public opinion and fueling disinformation. While existing research primarily focuses on identifying propaganda techniques, it lacks the ability to capture the broader motives and the impacts of such content. To address these challenges, we introduce propainsight, a conceptual framework grounded in foundational social science research, which systematically dissects propaganda into techniques, arousal appeals, and underlying intent. propainsight offers a more granular understanding of how propaganda operates across different contexts. Additionally, we present propagaze, a novel dataset that combines human-annotated data with high-quality synthetic data generated through a meticulously designed pipeline. Our experiments show that off-the-shelf LLMs struggle with propaganda analysis, but training with propagaze significantly improves performance. Fine-tuned Llama-7B-Chat achieves 203.4% higher text span IoU in technique identification and 66.2% higher BertScore in appeal analysis compared to 1-shot GPT-4-Turbo. Moreover, propagaze complements limited human-annotated data in data-sparse and cross-domain scenarios, showing its potential for comprehensive and generalizable propaganda analysis.

Ashraf Haddad, Najwa Aaraj, Preslav Nakov et al.

In recent years, a proliferation of cyber-security threats and diversity has been on the rise culminating in an increase in their reporting and analysis. To counter that, many non-profit organizations have emerged in this domain, such as MITRE and OSWAP, which have been actively tracking vulnerabilities, and publishing defense recommendations in standardized formats. As producing data in such formats manually is very time-consuming, there have been some proposals to automate the process. Unfortunately, a major obstacle to adopting supervised machine learning for this problem has been the lack of publicly available specialized datasets. Here, we aim to bridge this gap. In particular, we focus on mapping CVE records into MITRE CWE Weaknesses, and we release to the research community a manually annotated dataset of 4,012 records for this task. With a human-in-the-loop framework in mind, we approach the problem as a ranking task and aim to incorporate reinforced learning to make use of the human feedback in future work. Our experimental results using fine-tuned deep learning models, namely Sentence-BERT and rankT5, show sizable performance gains over BM25, BERT, and RoBERTa, which demonstrates the need for an architecture capable of good semantic understanding for this task.

Tariq Alhindi, Smaranda Muresan, Preslav Nakov

Recognizing fallacies is crucial for ensuring the quality and validity of arguments across various domains. However, computational fallacy recognition faces challenges due to the diverse genres, domains, and types of fallacies found in datasets. This leads to a highly multi-class, and even multi-label, setup with substantial class imbalance. In this study, we aim to enhance existing models for fallacy recognition by incorporating additional context and by leveraging large language models to generate synthetic data, thus increasing the representation of the infrequent classes. We experiment with GPT3.5 to generate synthetic examples and we examine the impact of prompt settings for this. Moreover, we explore zero-shot and few-shot scenarios to evaluate the effectiveness of using the generated examples for training smaller models within a unified fallacy recognition framework. Furthermore, we analyze the overlap between the synthetic data and existing fallacy datasets. Finally, we investigate the usefulness of providing supplementary context for detecting fallacy types that need such context, e.g., diversion fallacies. Our evaluation results demonstrate consistent improvements across fallacy types, datasets, and generators. The code and the synthetic datasets are all publicly available.

Gurpreet Gosal, Yishi Xu, Gokul Ramakrishnan et al.

We present an efficient method for adapting a monolingual Large Language Model (LLM) to another language, addressing challenges of catastrophic forgetting and tokenizer limitations. We focus this study on adapting Llama 2 to Arabic. Our two-stage approach begins with expanding the vocabulary and training only the embeddings matrix, followed by full model continual pre-training on a bilingual corpus. By continually pre-training on a mix of Arabic and English corpora, the model retains its proficiency in English while acquiring capabilities in Arabic. Our approach results in significant improvements in Arabic and slight enhancements in English, demonstrating cost-effective cross-lingual transfer. We perform ablations on embedding initialization techniques, data mix ratios, and learning rates and release a detailed training recipe. To demonstrate generalizability of this approach we also adapted Llama 3 8B to Arabic and Llama 2 13B to Hindi.

Luke Bates, Peter Ebert Christensen, Preslav Nakov et al.

Templatic memes, characterized by a semantic structure adaptable to the creator's intent, represent a significant yet underexplored area within meme processing literature. With the goal of establishing a new direction for computational meme analysis, here we create a knowledge base composed of more than 5,200 meme templates, information about them, and 54,000 examples of template instances (templatic memes). To investigate the semantic signal of meme templates, we show that we can match memes in datasets to base templates contained in our knowledge base with a distance-based lookup. To demonstrate the power of meme templates, we create TSplit, a method to reorganize datasets, where a template or templatic instance can only appear in either the training or test split. Our re-split datasets enhance general meme knowledge and improve sample efficiency, leading to more robust models. Our examination of meme templates results in state-of-the-art performance for every dataset we consider, paving the way for analysis grounded in templateness.

Rania Elbadry, Sarfraz Ahmad, Ahmed Heakl et al.

English financial NLP has progressed rapidly through benchmarks for sentiment, document understanding, and financial question answering, while Arabic financial NLP remains comparatively under-explored despite strong practical demand for trustworthy finance and Islamic-finance assistants. We introduce SAHM, a document-grounded benchmark and instruction-tuning dataset for Arabic financial NLP and Shari'ah-compliant reasoning. SAHM contains 14,380 expert-verified instances spanning seven tasks: AAOIFI standards QA, fatwa-based QA/MCQ, accounting and business exams, financial sentiment analysis, extractive summarization, and event-cause reasoning, curated from authentic regulatory, juristic, and corporate sources. We evaluate 19 strong open and proprietary LLMs using task-specific metrics and rubric-based scoring for open-ended outputs, and find that Arabic fluency does not reliably translate to evidence-grounded financial reasoning: models are substantially stronger on recognition-style tasks than on generation and causal reasoning, with the largest gaps on event-cause reasoning. We release the benchmark, evaluation framework, and an instruction-tuned model to support future research on trustworthy Arabic financial NLP.

Saptarshi Sengupta, Wenpeng Yin, Preslav Nakov et al.

In this paper, we investigate Extractive Question Answering (EQA) with Large Language Models (LLMs) under domain drift, i.e., can LLMs generalize to domains that require specific knowledge such as medicine and law in a zero-shot fashion without additional in-domain training? To this end, we devise a series of experiments to explain the performance gap empirically. Our findings suggest that: (a) LLMs struggle with dataset demands of closed domains such as retrieving long answer spans; (b) Certain LLMs, despite showing strong overall performance, display weaknesses in meeting basic requirements as discriminating between domain-specific senses of words which we link to pre-processing decisions; (c) Scaling model parameters is not always effective for cross domain generalization; and (d) Closed-domain datasets are quantitatively much different than open-domain EQA datasets and current LLMs struggle to deal with them. Our findings point out important directions for improving existing LLMs.

Rabindra Nath Nandi, Firoj Alam, Preslav Nakov

The spread of fake news, propaganda, misinformation, disinformation, and harmful content online raised concerns among social media platforms, government agencies, policymakers, and society as a whole. This is because such harmful or abusive content leads to several consequences to people such as physical, emotional, relational, and financial. Among different harmful content \textit{trolling-based} online content is one of them, where the idea is to post a message that is provocative, offensive, or menacing with an intent to mislead the audience. The content can be textual, visual, a combination of both, or a meme. In this study, we provide a comparative analysis of troll-based memes classification using the textual, visual, and multimodal content. We report several interesting findings in terms of code-mixed text, multimodal setting, and combining an additional dataset, which shows improvements over the majority baseline.