The proliferation of AI-Generated Content (AIGC), especially deepfake videos, poses a severe threat to social trust by enabling fraud, privacy violations and disinformation. Existing AI-generated video detection (AGVD) benchmarks focus on open-source model generated videos, yet commercial closed-source models produce more realistic, temporally coherent videos that are underexplored in detection research. To fill this gap, we present Chameleon, a commercial-grade dataset with 1,700 AI-generated videos from 600 real-world sources across three key domains (News, Speech, Recommendation), featuring high resolution, rich annotations and 3D consistency metrics for dynamic scene spatial coherence, shifting detection from face-centric forgery to holistic scene forensics. This benchmark assesses models on two core tasks: accurate AI video detection in real-world conditions and forensic backtracking of original sources. Experimental results reveal critical limitations of existing methods in detecting and backtracking high-fidelity, spatiotemporally consistent videos from commercial closed-source models, highlighting current methods’ flawed forensic reasoning and establishing Chameleon as a vital challenge for AIGC security research. The code and data are available at https://github.com/lxixim/Chameleon.

Abstract In the context of multi-scenario recommendation, multi-scenario click-through-rate (MS-CTR) prediction plays a crucial role in effectively personalizing recommendations on commercial platforms. However, when confronted with multi-scenario data, models often face challenges such as overfitting, inadequate feature representation, and unstable optimization, which hinder the performance and reliability of CTR prediction. To tackle these challenges, this study introduces the enhanced scene-aware transformer (ESAT) framework for MS-CTR prediction. This framework is divided into five modules. First, the structural position-aware scene encoding module converts scene attributes into fixed-dimensional embedding vectors and the scene adaptive transformation module uses a nonlinear transformation to dynamically adjust scene features. The cross-scene regularization module uses multi-sample dropout technology to enhance generalization ability and prevent overfitting. The scene-aware discriminative learning module applies contrastive learning to optimize the similarity between samples. Finally, the hierarchical stability control module introduces ClippyGrad optimizer and $L_\infty $ regularization to accurately control gradient updates, avoid excessive steps, and improve training stability. Experiments conducted on large-scale multi-scenario datasets confirm the effectiveness of the proposed module. The results indicate that the ESAT model substantially elevates performance in MS-CTR prediction tasks, especially in terms of generalization to new scenes, precision in feature representation, and stability in parameter updates.

Weixiong Zheng, Peijian Zeng, YiWei Li, Hongyan Wu, Nankai Lin, Junhao Chen, Aimin Yang, Yongmei Zhou. Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2025.

The performance of multilingual language models (MLLMs) is notably inferior for low-resource languages (LRL) compared to high-resource ones, primarily due to the limited available corpus during the pre-training phase. This inadequacy stems from the under-representation of low-resource language words in the subword vocabularies of MLLMs, leading to their misidentification as unknown or incorrectly concatenated subwords. Previous approaches are based on frequency sorting to select words for augmenting vocabularies. However, these methods overlook the fundamental disparities between model representation distributions and frequency distributions. To address this gap, we introduce a novel Entropy-Consistency Word Selection (ECWS) method, which integrates semantic and frequency metrics for vocabulary augmentation. Our results indicate an improvement in performance, supporting our approach as a viable means to enrich vocabularies inadequately represented in current MLLMs.

Chinese Semantic Error Recognition (CSER) has always been a weak link in Chinese language processing due to the complexity and obscureness of Chinese semantics. Existing research has gradually focused on leveraging pre-trained models to perform CSER. Although some researchers have attempted to integrate syntax information into the pre-trained language model, it requires training the models from scratch, which is time-consuming and laborious. Furthermore, despite the existence of datasets for CSER, the constrained size of these datasets impairs the performance of the models. Thus, in order to address the difficulty posed by a limited sample set and the need of annotating samples with semantic-level errors, we propose a Pseudo-label Data Construction method for CSER (PDC-CSER), generating pseudo-labels for augmented samples based on perplexity and model respectively, which overcomes the difficulty of constructing pseudo-label data containing semantic-level errors and ensures the quality of pseudo-labels. Moreover, we propose a CSER method with the Dependency Syntactic Attention mechanism (CSER-DSA) to explicitly infuse dependency syntactic information only in the fine-tuning stage, achieving robust performance, and simultaneously reducing substantial computing power and time cost. Results demonstrate that the pseudo-label technology PDC-CSER and the semantic error recognition method CSER-DSA surpass the existing models

Contrastive learning is a primary approach to mitigating data sparsity in sequential recommendation, but its improvements on the item side are limited due to data augmentation impairing user representations. To address these challenges, this paper introduces the Global Graph Attention for Contrastive Sequential Recommendation (GGACSR). GGACSR integrates graph neural networks and self-attention mechanisms, with an Attention Convolution Layer replacing nonlinear transformations in Graph Convolutional Networks (GCNs) with Q, K, and V vector operations, facilitating better handling of sequence dependencies. Leveraging a global user connection graph and projecting embeddings into a lower dimension effectively improves item and user representations. Overall, GGACSR outperforms existing baselines on three public datasets by more accurately capturing complex relationships and adapting user preferences.

Recently, Large Language Models (LLMs) have shown remarkable success across a variety of tasks, with rapid advancements in supporting multilingual capabilities. However, these models exhibit varying degrees of demographic biases in text classification tasks. Most existing research focuses on debiasing pre-trained models or addressing biases in monolingual text classification, resulting in limited exploration in multilingual contexts. To solve the above problems, this paper introduces a tWo-stAge in-conText learning dEbiasing fRamework (WATER). Our approach does not require updating the model's parameters and is adaptable to any language. It includes three key modules: sample selection, sample filtering, and template filling and prediction. In the first stage, we leverage a sample selection module to identify text that closely matches the model embeddings. In the second stage, we introduce an innovative Contextual Disparity Measure (CDM) in the sample filtering module to filter out samples that effectively address the bias associated with specific attributes. Finally, the template filling and prediction module is used to fill the selected samples into the template and input them into the model to complete the multilingual text classification task. Our experimental results verify the effectiveness of our method in mitigating biases related to four sensitive attributes of gender, age, race, and country, demonstrating its potential to improve the fairness and accuracy of LLMs in multilingual classification tasks.

Natural language processing models learn powerful language representation abilities from vast amounts of data, but they also inherit societal biases embedded in that data. Current research on debiasing often struggles to balance the removal of model bias with the preservation of model performance. Most existing approaches depend on fine-tuning model parameters, which can introduce uncertainties in model performance due to the modifications made to these parameters. In this paper, we propose a novel debiasing framework called FairTriplet. First, this framework employs prefix tuning to freeze the parameters of the original pre-trained model. Then, it optimizes the prefix parameters through two debiasing terms. These two debiasing terms function by reducing the semantic distance between social groups (e.g., male and female) and increasing the semantic distance between social groups and neutral attributes (e.g., family and occupation) in the semantic space. This approach not only removes bias from the model but also preserves its performance. Experimental results demonstrate that FairTriplet achieves state-of-the-art (SOTA) levels in debiasing while maintaining model performance on GLUE downstream tasks.

Cross-lingual Aspect-based Sentiment Analysis (XABSA) aims to extract aspect-level sentiments across multiple languages. This task typically relies on source language data to train models and transfer them to target languages, so it faces significant challenges such as data scarcity and language disparities. To this end, this study proposes a code-Mixed In-conteXt lEaRning (MIXER). We design four kinds of demonstration retrieval libraries to introduce Code-mixed In-Context Demonstrations (CICD), which use the code-mixed mechanism to integrate the features of the target language and enrich the target language's knowledge while retaining the source language's knowledge. Language-Specific tags (LST) are introduced to enhance the model's understanding of multilingual demonstrations. To validate the effectiveness of MIXER, we conduct extensive experiments on the SemEval-2016 dataset, comparing its performance against existing XABSA methods. The experimental results show that MIXER performs better than existing XABSA methods with average F1 scores on the Mistral and Llama3 improved by 1.59% and 1.44%, respectively, highlighting its potential for broader multilingual applications.

Industrial Anomaly Detection (IAD) is critical for ensuring product quality by identifying defects. Traditional methods such as feature embedding and reconstruction-based approaches require large datasets and struggle with scalability. Existing vision-language models (VLMs) and Multimodal Large Language Models (MLLMs) address some limitations but rely on mask annotations, leading to high implementation costs and false positives. Additionally, industrial datasets like MVTec-AD and VisA suffer from severe class imbalance, with defect samples constituting only 23.8 % and 11.1 % of total data respectively. To address these challenges, we propose a reward function that dynamically prioritizes rare defect patterns during training to handle class imbalance. We also introduce a mask-free reasoning framework using Chain of Thought (CoT) and Group Relative Policy Optimization (GRPO) mechanisms, enabling anomaly detection directly from raw images without annotated masks. This approach generates interpretable step-by-step explanations for defect localization. Our method achieves state-of-the-art performance, outperforming prior approaches by 36% in accuracy on MVTec-AD and 16% on VisA. By eliminating mask dependency and reducing costs while providing explainable outputs, this work advances industrial anomaly detection and supports scalable quality control in manufacturing.

Testing conditional independence is a critical task, particularly in causal discovery and learning in Bayesian networks. However, in many real-world scenarios, variables are often measured with errors, such as those introduced by insufficient measurement accuracy, complicating the testing process. This paper focuses on testing conditional independence in the linear non-Gaussian measurement error model, under the condition that measurement error noise follows a Gaussian distribution. By leveraging high-order cumulants, we derive rank constraints on the cumulant matrix and establish their role in effectively assessing conditional independence, even in the presence of measurement errors. Based on these theoretical results, we leverage the rank constraints of the cumulant matrix as a tool for conditional independence testing and incorporate it into the PC algorithm, resulting in the PC-ME algorithm — a method designed to learn causal structures from observed data while accounting for measurement errors. Experimental results demonstrate that the proposed method outperforms existing approaches, particularly in cases other methods encounter difficulties.

The primary challenge in text-video retrieval lies in achieving cross-modal semantic alignment, particularly the discrepancy between the conciseness of textual descriptions, which often fail to fully encapsulate the breadth of video content, and the redundancy in video data, which introduces noise and masks important semantic features. Current methods align text and video by mapping them into a shared feature space. Despite notable advancements, the inherent differences in modality-specific representations create a bottleneck for fixed-point embedding techniques, making models highly sensitive to dataset distribution and hindering their generalization ability. In this paper, we present DomainDiff, a framework that enhances the embedding space through a two-stage process. In the first stage, stochastic domain modeling, we semantically expand text embeddings to explore potential regions aligned with video content. Simultaneously, we filter video segments to reduce redundancy and highlight key frames. In the second stage, the dynamic agent attention diffusion network, we leverage the generative properties of diffusion models to optimize the embedding space by viewing it from a joint probability distribution perspective. An agent attention mechanism dynamically integrates text and video features, ensuring accurate cross-modal alignment. Experimental results demonstrate that DomainDiff significantly improves retrieval performance across five benchmark datasets, with R@1 improvements ranging from 3% to 7.4%. Moreover, DomainDiff outperforms existing methods in handling long videos and complex textual descriptions, showcasing superior semantic robustness and generalization across varying distributions.

Molecular retrieval is critical in drug discovery and molecular design. Traditional discriminative methods often model the conditional probability distribution of retrieving candidates, treating the query text as a deterministic input. However, these approaches have notable limitations: (1) They often overlook the statistical properties of the original data distributions of queries and candidates, preventing the recognition of out-of-distribution data. (2) They struggle to balance retrieval accuracy and diversity when processing open-ended semantic queries. To address these challenges, we introduce DiffTMR, a novel framework that reformulates text-molecule retrieval as a reverse denoising process, progressively generating the joint distribution of candidates and queries from noises. DiffTMR uniquely integrates hierarchical diffusion alignment with dynamic perturbation embedding mechanisms. By employing text-anchored perturbations, it enhances the diversity of molecular representations, and through global-local progressive denoising, it achieves cross-modal hierarchical alignment. This leads to significant improvements in retrieval accuracy and out-of-domain generalization. Evaluations on benchmark datasets ChEBI-20 and PCdes demonstrate that DiffTMR surpasses current leading baselines by 4.2%-5.4% in Hits@1 metrics and exhibits superior performance in out-of-domain retrieval tasks.

Abstract Data augmentation-based contrastive learning has been successfully employed in Variational AutoEncoders sequence recommendation systems to tackle the issue of data sparsity. Nevertheless, this strategy is generally less advantageous for tail users. The prospective transmission of information from head-to-tail users to alleviate long-tail impact is encouraging. However, data augmentation distorts the original sequence and embeds stochastic noise into latent variables, impeding the decoder’s capacity to accurately identify the user’s true preferences. In addition, contrastive learning seeks to achieve consistency in the latent variables of both the original and augmented data. However, the presence of noise in the augmented data might hamper the encoding of latent variables from the original data, especially impacting head users. In order to address these challenges, this work introduces a new sequence recommendation model called the Filter-enhanced Contrastive Variational Autoencoder (FeCVAE). It employs Fourier filters and adversarial attack training to minimize the impact of stochastic noise, thereby improving the quality of latent variables and facilitating more accurate decoder outputs. Moreover, a user enhancer is introduced to leverage knowledge from head users to empower tail users, thereby alleviating the long-tail effect. The efficacy of FeCVAE is demonstrated through comprehensive experiments across four benchmark datasets.

Abstract Curriculum learning (CL) is a training strategy that imitates how humans learn, by gradually introducing more complex samples and information to the model. However, in multi-label emotion classification (MEC) tasks, using a traditional CL approach can result in overfitting on easy samples and lead to biased training. Additionally, the sample difficulty varies as the model trains. To address these challenges, we propose a novel CL framework for MEC tasks called CLF-MEC. Unlike traditional approaches that assess difficulty at the sample level, we utilize category-level assessment to determine the difficulty level of samples. As the model identifies a category well, the score for that category’s samples is reduced, ensuring dynamic changes in the sample difficulty are accounted for. Our CL framework employs two training modes, namely “learning” and “tackling.” These two processes are trained alternatively to imitate the “learning-tackling” process in human learning. This ensures that samples from hard-to-learn categories receive more attention. During the “tackling” process, our method transforms the task of dealing with hard samples into an “easy” learning task by utilizing contrastive learning to enhance the semantic representation of those hard samples. Experimental results demonstrate that our CLF-MEC framework has achieved significant improvements in MEC.

Currently, the majority of research in grammatical error correction (GEC) is concentrated on universal languages, such as English and Chinese. Many low-resource languages lack accessible evaluation corpora. How to efficiently construct high-quality evaluation corpora for GEC in low-resource languages has become a significant challenge. To fill these gaps, in this article, we present a framework for constructing GEC corpora. Specifically, we focus on Indonesian as our research language and construct an evaluation corpus for Indonesian GEC using the proposed framework, addressing the limitations of existing evaluation corpora in Indonesian. Furthermore, we investigate the feasibility of utilizing existing large language models (LLMs), such as GPT-3.5-Turbo and GPT-4, to streamline corpus annotation efforts in GEC tasks. The results demonstrate significant potential for enhancing the performance of LLMs in low-resource language settings. Our code and corpus can be obtained from https://github.com/GKLMIP/GEC-Construction-Framework .

The prevalence of fake news online has become a significant societal concern. To combat this, multimodal detection techniques based on images and text have shown promise. Yet, these methods struggle to analyze complex relationships within and between modalities due to the diverse discriminative elements in the news content. In addition, research on multimodal and multi-class fake news detection remains insufficient. To address the above challenges, in this article, we propose a novel detection model, GS 2 F, leveraging g raph s tructure and g uided s emantic f usion. Specifically, we construct a multimodal graph structure to align two modalities and employ graph contrastive learning for refined fusion representations. Furthermore, a guided semantic fusion module is introduced to maximize the utilization of single-modal information and a dynamic contribution assignment layer is designed to weigh the importance of image, text, and multimodal features. Experimental results on Fakeddit demonstrate that our model outperforms existing methods, marking a step forward in the multimodal and multi-class fake news detection.

Sequential recommendation plays a key role in recommender systems, where the goal is to predict a user’s future points of interest by analyzing his or her historical interactions. This process not only requires the system to be able to accurately identify and recommend items that are likely to be of interest to the user but also ensures that all items receive equal exposure to prevent over-concentration or marginalization of items due to algorithmic bias. To address these challenges, in this paper, we propose a novel Adversarial Debiasing framework based on Sinusoidal Embedding for sequential recommendation, ADSE. This framework employs sinusoidal position embeddings to extract positional information between sequences more precisely and utilizes a dropout strategy to optimize the handling of cold-start sequences, aiming to resolve the cold-start issue while maintaining the semantics of the original sequences. Additionally, adversarial training was incorporated to reduce implicit bias due to assuming interactions in the calculation of exposure.

The task of retrieving audio content relevant to lyric queries and vice versa plays a critical role in music-oriented applications. In this process, robust feature representations have to be learned for two modalities. Furthermore, interactions between different modalities should be properly captured at a fine-grained level. Existing approaches can effectively extract modal representations and perform retrieving between different modalities through alignment. However, these approaches model interactions between audio and lyrics in a coarse-grained manner. Especially the input features and interactions between enhanced representations produced by the alignment module are largely ignored, resulting in low-quality modality representations for final retrieval. This paper presents a novel method named CMRF that accomplishes cross-modal interactions via a reinforcement feedback procedure to learn high-quality multi-modal embeddings. Initially, we implicitly assimilate representations across distinct modalities via directional pairwise cross-modal attention. Subsequently, our approach recurrently identifies pivotal constituents within these elevated-level attributes to engage with the primary input features via reinforcement learning, thus augmenting the quality of multi-modal embeddings. In addition, we introduce a novel audio-lyrics datasetAL-song, which consists of paired audio with corresponding lyrics for the audio-lyrics retrieval task. The empirical findings derived from theAL-songdataset and the benchmark datasetSounddescssubstantiate the efficacy and efficiency of CMRF when juxtaposed with state-of-the-art methodologies.

Aspect-based sentiment analysis (ABSA), an extensively researched area in the field of natural language processing (NLP), predicts the sentiment expressed in a text relative to the corresponding aspect. Unfortunately, most languages lack sufficient annotation resources; thus, an increasing number of recent researchers have focused on cross-lingual aspect-based sentiment analysis (XABSA). However, most recent studies focus only on cross-lingual data alignment instead of model alignment. Therefore, we propose a novel framework, CL-XABSA: contrastive learning for cross-lingual aspect-based sentiment analysis. Based on contrastive learning, we close the distance between samples with the same label in different semantic spaces, achieving convergence of semantic spaces of different languages. Specifically, we design two contrastive objectives, token-level contrastive learning of token embeddings (TL-CTE) and sentiment-level contrastive learning of token embeddings (SL-CTE), to unify the semantic space of source and target languages. Since CL-XABSA can receive datasets in multiple languages during training, it can be further extended to multilingual aspect-based sentiment analysis (MABSA). To further improve the model performance, we perform knowledge distillation with target-language unlabeled data. In the distillation XABSA task, we further explore the effectiveness of different data (source dataset, translated dataset, and code-switched dataset). The results demonstrate that the proposed method has a certain improvement in the three XABSA tasks, distillation XABSA and MABSA. The source code of this paper is publicly available athttps://github.com/GKLMIP/CL-XABSA.

The goal of cross-domain sentiment classification is to utilise useful information in the source domain to help classify sentiment polarity in the target domain, which has a large number of unlabelled data. Most of the existing methods focus on extracting the invariant features between two domains. But they cannot make better use of the unlabelled data in the target domain. To solve this problem, we present a deep transfer learning mechanism (DTLM) for fine-grained cross-domain sentiment classification. DTLM provides a transfer mechanism to better transfer sentiment across domains by incorporating BERT(Bidirextional Encoder Representations from Transformers) and KL (Kullback-Leibler) divergence. We introduce BERT as a feature encoder to map the text data of different domains into a shared feature space. Then, we design a domain adaptive model using KL divergence to eliminate the difference of feature distribution between the source domain and target domain. In addition, we introduce the entropy minimisation and consistency regularisation to process unlabelled samples in the target domain. Extensive experiments on the datasets from YelpAspect, SemEval 2014 task 4 and Twitter not only demonstrate the effectiveness of our proposed method but also provide a better way for cross-domain sentiment classification.