We introduce a new task called Defeasible Visual Entailment (DVE), where the goal is to allow the modification of the entailment relationship between an image premise and a text hypothesis based on an additional update. While this concept is well-established in Natural Language Inference, it remains unexplored in visual entailment. At a high level, DVE enables models to refine their initial interpretations, leading to improved accuracy and reliability in various applications such as detecting misleading information in images, enhancing visual question answering, and refining decision-making processes in autonomous systems. Existing metrics do not adequately capture the change in the entailment relationship brought by updates. To address this, we propose a novel inference-aware evaluator designed to capture changes in entailment strength induced by updates, using pairwise contrastive learning and categorical information learning. Additionally, we introduce a reward-driven update optimization method to further enhance the quality of updates generated by multimodal models. Experimental results demonstrate the effectiveness of our proposed evaluator and optimization method.

Sarcasm is a complex linguistic phenomenon that involves a disparity between literal and intended meanings, making it challenging for sentiment analysis and other emotion-sensitive tasks. While traditional sarcasm detection methods primarily focus on text, recent approaches have incorporated multimodal information. However, the application of Large Visual Language Models (LVLMs) in Multimodal Sarcasm Analysis (MSA) remains underexplored. In this paper, we evaluate LVLMs in MSA tasks, specifically focusing on Multimodal Sarcasm Detection and Multimodal Sarcasm Explanation. Through comprehensive experiments, we identify key limitations, such as insufficient visual understanding and a lack of conceptual knowledge. To address these issues, we propose a training-free framework that integrates in-depth object extraction and external conceptual knowledge to improve the model's ability to interpret and explain sarcasm in multimodal contexts. The experimental results on multiple models show the effectiveness of our proposed framework. The code is available at https://github.com/cp-cp/LVLM-MSA.

Shuo Yan, Ruochen Li, Ziming Luo, Zimu Wang, Daoyang Li, Liqiang Jing, Kaiyu He, Peilin Wu, Juntong Ni, George Michalopoulos, Yue Zhang, Ziyang Zhang, Mian Zhang, Zhiyu Chen, Xinya Du. Proceedings of the 2025 Conference on Empirical Methods in Natural Language Processing. 2025.

Large Language Models (LLMs) and Large Vision-Language Models (LVLMs) have demonstrated impressive language/vision reasoning abilities, igniting the recent trend of building agents for targeted applications such as shopping assistants or AI software engineers. Recently, many data science benchmarks have been proposed to investigate their performance in the data science domain. However, existing data science benchmarks still fall short when compared to real-world data science applications due to their simplified settings. To bridge this gap, we introduce DSBench, a comprehensive benchmark designed to evaluate data science agents with realistic tasks. This benchmark includes 466 data analysis tasks and 74 data modeling tasks, sourced from Eloquence and Kaggle competitions. DSBench offers a realistic setting by encompassing long contexts, multimodal task backgrounds, reasoning with large data files and multi-table structures, and performing end-to-end data modeling tasks. Our evaluation of state-of-the-art LLMs, LVLMs, and agents shows that they struggle with most tasks, with the best agent solving only 34.12% of data analysis tasks and achieving a 34.74% Relative Performance Gap (RPG). These findings underscore the need for further advancements in developing more practical, intelligent, and autonomous data science agents.

Multimodal Large Language Models (MLLMs), Large Vision-Language Models (LVLMs), and Large Language Models (LLMs) have demonstrated remarkable capabilities across various tasks, including text-based reasoning and multimodal content generation. However, these models frequently generate hallucinations-factually incorrect or misleading content-that pose significant challenges, particularly in high-stakes domains such as healthcare, law, and finance. This tutorial provides a comprehensive exploration of hallucinations in MLLMs, LVLMs, and LLMs, examining their causes, detection methods, and mitigation strategies. We discuss different types of hallucination evaluation and benchmarking and explore state-of-the-art techniques for hallucination mitigation.

Sarcasm Explanation in Dialogue (SED) is a new yet challenging task, which aims to generate a natural language explanation for the given sarcastic dialogue that involves multiple modalities (i.e.,utterance, video, and audio). Although existing studies have achieved great success based on the generative pretrained language model BART, they overlook exploiting the sentiments residing in the utterance, video and audio, which play important roles in reflecting sarcasm that essentially involves subtle sentiment contrasts. Nevertheless, it is non-trivial to incorporate sentiments for boosting SED performance, due to three main challenges: 1) diverse effects of utterance tokens on sentiments; 2) gap between video-audio sentiment signals and the embedding space of BART; and 3) various relations among utterances, utterance sentiments, and video-audio sentiments. To tackle these challenges, we propose a novel sEntiment-enhanceD Graph-based multimodal sarcasm Explanation framework, named EDGE. In particular, we first propose a lexicon-guided utterance sentiment inference module, where a heuristic utterance sentiment refinement strategy is devised. We then develop a module named Joint Cross Attention-based Sentiment Inference (JCA-SI) by extending the multimodal sentiment analysis model JCA to derive the joint sentiment label for each video-audio clip. Thereafter, we devise a context-sentiment graph to comprehensively model the semantic relations among the utterances, utterance sentiments, and video-audio sentiments, to facilitate sarcasm explanation generation. Extensive experiments on the publicly released dataset WITS verify the superiority of our model over cutting-edge methods.

Despite commendable achievements made by existing work, prevailing multimodal sarcasm detection studies rely more on textual content over visual information. It unavoidably induces spurious correlations between textual words and labels, thereby significantly hindering the models' generalization capability. To address this problem, we define the task of out-of-distribution (OOD) multimodal sarcasm detection, which aims to evaluate models' generalizability when the word distribution is different in training and testing settings. Moreover, we propose a novel debiasing multimodal sarcasm detection framework with contrastive learning, which aims to mitigate the harmful effect of biased textual factors for robust OOD generalization. In particular, we first design counterfactual data augmentation to construct the positive samples with dissimilar word biases and negative samples with similar word biases. Subsequently, we devise an adapted debiasing contrastive learning mechanism to empower the model to learn robust task-relevant features and alleviate the adverse effect of biased words. Extensive experiments show the superiority of the proposed framework.

Existing sign language translation methods follow a two-stage pipeline: first converting the sign language video to a gloss sequence (i.e., Sign2Gloss) and then translating the generated gloss sequence into a spoken language sentence (i.e., Gloss2Text). While previous studies have focused on boosting the performance of the Sign2Gloss stage, we emphasize the optimization of the Gloss2Text stage. However, this task is non-trivial due to two distinct features of Gloss2Text: (1) isolated gloss input and (2) low-capacity gloss vocabulary. To address these issues, we propose a vision and context knowledge enhanced Gloss2Text model, named VK-G2T, which leverages the visual content of the sign language video to learn the properties of the target sentence and exploit the context knowledge to facilitate the adaptive translation of gloss words. Extensive experiments conducted on a Chinese benchmark validate the superiority of our model.

Text response generation for multimodal task-oriented dialog systems, which aims to generate the proper text response given the multimodal context, is an essential yet challenging task. Although existing efforts have achieved compelling success, they still suffer from two pivotal limitations: (1) overlook the benefit of generative pretraining and (2) ignore the textual context-related knowledge . To address these limitations, we propose a novel dual knowledge-enhanced generative pretrained language mode for multimodal task-oriented dialog systems (DKMD), consisting of three key components: dual knowledge selection , dual knowledge-enhanced context learning , and knowledge-enhanced response generation . To be specific, the dual knowledge selection component aims to select the related knowledge according to both textual and visual modalities of the given context. Thereafter, the dual knowledge-enhanced context learning component targets seamlessly, integrating the selected knowledge into the multimodal context learning from both global and local perspectives, where the cross-modal semantic relation is also explored. Moreover, the knowledge-enhanced response generation component comprises a revised BART decoder, where an additional dot-product knowledge-decoder attention sub-layer is introduced for explicitly utilizing the knowledge to advance the text response generation. Extensive experiments on a public dataset verify the superiority of the proposed DKMD over state-of-the-art competitors.

Existing data-to-text generation efforts mainly focus on generating a coherent text from non-linguistic input data, such as tables and attribute–value pairs, but overlook that different application scenarios may require texts of different styles. Inspired by this, we define a new task, namely stylized data-to-text generation, whose aim is to generate coherent text for the given non-linguistic data according to a specific style. This task is non-trivial, due to three challenges: the logic of the generated text, unstructured style reference and biased training samples. To address these challenges, we propose a novel stylized data-to-text generation model, named StyleD2T, comprising three components: logic planning-enhanced data embedding, mask-based style embedding, and unbiased stylized text generation. In the first component, we introduce a graph-guided logic planner for attribute organization to ensure the logic of generated text. In the second component, we devise feature-level mask-based style embedding to extract the essential style signal from the given unstructured style reference. In the last one, pseudo triplet augmentation is utilized to achieve unbiased text generation, and a multi-condition based confidence assignment function is designed to ensure the quality of pseudo samples. Extensive experiments on a newly collected dataset from Taobao 1 have been conducted, and the results show the superiority of our model over existing methods.

Sarcasm is a sophisticated linguistic phenomenon that is prevalent on today's social media platforms. Multi-modal sarcasm detection aims to identify whether a given sample with multi-modal information (i.e., text and image) is sarcastic. This task's key lies in capturing both inter- and intra-modal incongruities within the same context. Although existing methods have achieved compelling success, they are disturbed by irrelevant information extracted from the whole image and text, or overlooking some important information due to the incomplete input. To address these limitations, we propose a Mutual-enhanced Incongruity Learning Network for multi-modal sarcasm detection, named MILNet. In particular, we design a local semantic-guided incongruity learning module and a global incongruity learning module. Moreover, we introduce a mutual enhancement module to take advantage of the underlying consistency between the two modules to boost the performance. Extensive experiments on a widely-used dataset demonstrate the superiority of our model over cutting-edge methods.

Multimodal Sarcasm Explanation (MuSE) is a new yet challenging task, which aims to generate a natural language sentence for a multimodal social post (an image as well as its caption) to explain why it contains sarcasm.Although the existing pioneer study has achieved great success with the BART backbone, it overlooks the gap between the visual feature space and the decoder semantic space, the objectlevel metadata of the image, as well as the potential external knowledge.To solve these limitations, in this work, we propose a novel mulTisource sEmantic grAph-based Multimodal sarcasm explanation scheme, named TEAM.In particular, TEAM extracts the object-level semantic meta-data instead of the traditional global visual features from the input image.Meanwhile, TEAM resorts to ConceptNet to obtain the external related knowledge concepts for the input text and the extracted object metadata.Thereafter, TEAM introduces a multisource semantic graph that comprehensively characterize the multi-source (i.e., caption, object meta-data, external knowledge) semantic relations to facilitate the sarcasm reasoning.Extensive experiments on a public released dataset MORE verify the superiority of our model over cutting-edge methods.

Existing work on Multimodal Sentiment Analysis (MSA) utilizes multimodal information for prediction yet unavoidably suffers from fitting the spurious correlations between multimodal features and sentiment labels. For example, if most videos with a blue background have positive labels in a dataset, the model will rely on such correlations for prediction, while "blue background'' is not a sentiment-related feature. To address this problem, we define a general debiasing MSA task, which aims to enhance the Out-Of-Distribution (OOD) generalization ability of MSA models by reducing their reliance on spurious correlations. To this end, we propose a general debiasing framework based on Inverse Probability Weighting (IPW), which adaptively assigns small weights to the samples with larger bias (i.e., the severer spurious correlations). The key to this debiasing framework is to estimate the bias of each sample, which is achieved by two steps: 1) disentangling the robust features and biased features in each modality, and 2) utilizing the biased features to estimate the bias. Finally, we employ IPW to reduce the effects of large-biased samples, facilitating robust feature learning for sentiment prediction. To examine the model's generalization ability, we keep the original testing sets on two benchmarks and additionally construct multiple unimodal and multimodal OOD testing sets. The empirical results demonstrate the superior generalization ability of our proposed framework. We have released the code to facilitate the reproduction https://github.com/Teng-Sun/GEAR.

Multimodal sentence summarization, aiming to generate a brief summary of the source sentence and image, is a new yet challenging task. Although existing methods have achieved compelling success, they still suffer from two key limitations: 1) lacking the adaptation of generative pre-trained language models for open-domain MMSS, and 2) lacking the explicit critical information modeling. To address these limitations, we propose a BART-MMSS framework, where BART is adopted as the backbone. To be specific, we propose a prompt-guided image encoding module to extract the source image feature. It leverages several soft to-be-learned prompts for image patch embedding, which facilitates the visual content injection to BART for open-domain MMSS tasks. Thereafter, we devise an explicit source critical token learning module to directly capture the critical tokens of the source sentence with the reference of the source image, where we incorporate explicit supervision to improve performance. Extensive experiments on a public dataset fully validate the superiority of our proposed method. In addition, the predicted tokens by the vision-guided key-token highlighting module can be easily understood by humans and hence improve the interpretability of our model.

In the context of COVID-19, numerous people present their opinions through social networks. It is thus highly desired to conduct sentiment analysis towards COVID-19 tweets to learn the public's attitudes, and facilitate the government to make proper guidelines for avoiding the social unrest. Although many efforts have studied the text-based sentiment classification from various domains (e.g., delivery and shopping reviews), it is hard to directly use these classifiers for the sentiment analysis towards COVID-19 tweets due to the domain gap. In fact, developing the sentiment classifier for COVID-19 tweets is mainly challenged by the limited annotated training dataset, as well as the diverse and informal expressions of user-generated posts. To address these challenges, we construct a large-scale COVID-19 dataset from Weibo and propose a dual COnsistency-enhanced semi-superVIseD network for Sentiment Anlaysis (COVID-SA). In particular, we first introduce a knowledge-based augmentation method to augment data and enhance the model's robustness. We then employ BERT as the text encoder backbone for both labeled data, unlabeled data, and augmented data. Moreover, we propose a dual consistency (i.e., label-oriented consistency and instance-oriented consistency) regularization to promote the model performance. Extensive experiments on our self-constructed dataset and three public datasets show the superiority of COVID-SA over state-of-the-art baselines on various applications.

Existing studies on multimodal sentiment analysis heavily rely on textual modality and unavoidably induce the spurious correlations between textual words and sentiment labels. This greatly hinders the model generalization ability. To address this problem, we define the task of out-of-distribution (OOD) multimodal sentiment analysis. This task aims to estimate and mitigate the bad effect of textual modality for strong OOD generalization. To this end, we embrace causal inference, which inspects the causal relationships via a causal graph. From the graph, we find that the spurious correlations are attributed to the direct effect of textual modality on the model prediction while the indirect one is more reliable by considering multimodal semantics. Inspired by this, we devise a model-agnostic counterfactual framework for multimodal sentiment analysis, which captures the direct effect of textual modality via an extra text model and estimates the indirect one by a multimodal model. During the inference, we first estimate the direct effect by the counterfactual inference, and then subtract it from the total effect of all modalities to obtain the indirect effect for reliable prediction. Extensive experiments show the superior effectiveness and generalization ability of our proposed framework.

Multi-modal Product Summary Generation is a new yet challenging task, which aims to generate a concise and readable summary for a product given its multi-modal content, e.g., its long text description and image. Although existing methods have achieved great success, they still suffer from three key limitations: 1) overlook the benefit of pre-training, 2) lack the representation-level supervision, and 3) ignore the diversity of the seller-generated data. To address these limitations, in this work, we propose a Vision-to-Prompt based multi-modal product summary generation framework, dubbed as V2P, where a Generative Pre-trained Language Model (GPLM) is adopted as the backbone. In particular, to maintain the original text capability of the GPLM and fully utilize the high-level concepts contained in the product image, we design V2P with two key components: vision-based prominent attribute prediction, and attribute prompt-guided summary generation. The first component works on obtaining the vital semantic attributes of the product from its image by the Swin Transformer, while the second component aims to generate the summary based on the product's long text description and the attribute prompts yielded by the first component with a GPLM. Towards comprehensive supervision over the second component, apart from the conventional output-level supervision, we introduce the representation-level regularization. Meanwhile, we design the data augmentation-based robustness regularization to handle the diverse inputs and improve the robustness of the second component. Extensive experiments on a large-scale Chinese dataset verify the superiority of our model over cutting-edge methods.