Emojis are globally used non-verbal cues in digital communication, and extensive research has examined how large language models (LLMs) understand and utilize emojis across contexts. While usually associated with friendliness or playfulness, it is observed that emojis may trigger toxic content generation in LLMs. Motivated by such a observation, we aim to investigate: (1) whether emojis can clearly enhance the toxicity generation in LLMs and (2) how to interpret this phenomenon.* We begin with a comprehensive exploration of emoji-triggered LLM toxicity generation by automating the construction of prompts with emojis to subtly express toxic intent. Experiments across 5 mainstream languages on 7 famous LLMs along with jailbreak tasks demonstrate that prompts with emojis could easily induce toxicity generation. To understand this phenomenon, we conduct model-level interpretations spanning semantic cognition, sequence generation and tokenization, suggesting that emojis can act as a heterogeneous semantic channel to bypass the safety mechanisms. To pursue deeper insights, we further probe the pre-training corpus and uncover potential correlation between the emoji-related data polution with the toxicity generation behaviors.

Renmiao Chen, Yida Lu, Shiyao Cui, Xuan Ouyang, Victor Shea-Jay Huang, Shumin Zhang, Chengwei Pan, Han Qiu, Minlie Huang. Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2026.

Shiyao Cui, QingLin Zhang, Di Wang, Yida Lu, Zhexin Zhang, Jinhua Gao, Jinglin Yang, Min He, Han Qiu, Minlie Huang. Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2026.

Junxiao Yang, Haoran Liu, Jinzhe Tu, Jiale Cheng, Zhexin Zhang, Shiyao Cui, Jiaqi Weng, Jialing Tao, Hui Xue, Hongning Wang, Han Qiu, Minlie Huang. Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2026.

Zhexin Zhang, Xian Qi Loye, Victor Shea-Jay Huang, Junxiao Yang, Qi Zhu, Shiyao Cui, Fei Mi, Lifeng Shang, Yingkang Wang, Hongning Wang, Minlie Huang. Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2026.

Yida Lu, Jiale Cheng, Zhexin Zhang, Shiyao Cui, Cunxiang Wang, Xiaotao Gu, Yuxiao Dong, Jie Tang, Hongning Wang, Minlie Huang. Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2025.

Jailbreaking attacks can effectively induce unsafe behaviors in Large Language Models (LLMs); however, the transferability of these attacks across different models remains limited. This study aims to understand and enhance the transferability of gradient-based jailbreaking methods, which are among the standard approaches for attacking white-box models. Through a detailed analysis of the optimization process, we introduce a novel conceptual framework to elucidate transferability and identify superfluous constraints—specifically, the response pattern constraint and the token tail constraint—as significant barriers to improved transferability. Removing these unnecessary constraints substantially enhances the transferability and controllability of gradient-based attacks. Evaluated on Llama-3-8B-Instruct as the source model, our method increases the overall Transfer Attack Success Rate (T-ASR) across a set of target models with varying safety levels from 18.4% to 50.3%, while also improving the stability and controllability of jailbreak behaviors on both source and target models.

Jailbreak attacks against multimodal large language Models (MLLMs) are a significant research focus. Current research predominantly focuses on maximizing attack success rate (ASR), often overlooking whether the generated responses actually fulfill the attacker's malicious intent. This oversight frequently leads to low-quality outputs that, while successful in bypassing safety filters, lack substantial harmful content. To address this gap, we propose JPS, Jailbreak MLLMs with collaborative visual Perturbation and textual Steering, which achieves jailbreaks via corporation of visual image and textually steering prompt. Specifically, JPS utilizes target-guided adversarial image perturbations for effective safety bypass, complemented by ''steering prompt'' optimized via a multi-agent system to specifically guide LLM responses fulfilling the attackers' intent. These visual and textual components undergo iterative co-optimization for enhanced performance. To evaluate the quality of attack outcomes, we propose the Malicious Intent Fulfillment Rate (MIFR) metric, assessed using a Reasoning-LLM-based evaluator. Our experiments show JPS sets a new state-of-the-art in both ASR and MIFR across various MLLMs and benchmarks, with analyses confirming its efficacy. Codes are available at https://github.com/thu-coai/JPS Warning: This paper contains potentially sensitive contents.

Toxicity detection in multimodal text-image content faces growing challenges, especially with multimodal implicit toxicity, where each modality appears benign on its own but conveys hazard when combined. Multimodal implicit toxicity appears not only as formal statements in social platforms but also prompts that can lead to toxic dialogs from Large Vision-Language Models (LVLMs). Despite the success in unimodal text or image moderation, toxicity detection for multimodal content, particularly the multimodal implicit toxicity, remains underexplored. To fill this gap, we comprehensively build a taxonomy for multimodal implicit toxicity (MMIT) and introduce an MMIT-dataset, comprising 2,100 multimodal statements and prompts across 7 risk categories (31 sub-categories) and 5 typical cross-modal correlation modes. To advance the detection of multimodal implicit toxicity, we build ShieldVLM, a model which identifies implicit toxicity in multimodal statements, prompts and dialogs via deliberative cross-modal reasoning. Experiments show that ShieldVLM outperforms existing strong baselines in detecting both implicit and explicit toxicity. The model and dataset will be publicly available to support future researches (Warning: This paper contains potentially sensitive contents). Warning: This paper contains potentially sensitive contents.

Yilong Chen, Junyuan Shang, Zhenyu Zhang, Shiyao Cui, Tingwen Liu, Shuohuan Wang, Yu Sun, Hua Wu. Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2024.

Yilong Chen, Guoxia Wang, Junyuan Shang, Shiyao Cui, Zhenyu Zhang, Tingwen Liu, Shuohuan Wang, Yu Sun, Dianhai Yu, Hua Wu. Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2024.

With the increasing availability of multimodal content on social media, consisting primarily of text and images, multimodal named entity recognition (MNER) has gained a wide-spread attention. A fundamental challenge of MNER lies in effectively aligning different modalities. However, the majority of current approaches rely on word-based sequence labeling framework and align the image and text at inconsistent semantic levels (whole image-words or regions-words). This misalignment may lead to inferior entity recognition performance. To address this issue, we propose an effective span-based method, named SMNER, which achieves a more consistent multimodal alignment from the perspectives of information-theoretic and cross-modal interaction, respectively. Specifically, we first introduce a cross-modal information bottleneck module for the global-level multimodal alignment (whole image-whole text). This module aims to encourage the semantic distribution of the image to be closer to the semantic distribution of the text, which can enable the filtering out of visual noise. Next, we introduce a cross-modal attention module for the local-level multimodal alignment (regions-spans), which captures the correlations between regions in the image and spans in the text, enabling a more precise alignment of the two modalities. Extensive ex- periments conducted on two benchmark datasets demonstrate that SMNER outperforms the state-of-the-art baselines.

Chinese Spelling Correction (CSC) aims to detect and correct the misspelled characters in Chinese texts. Recent studies have achieved great success by incorporating the phonetic information for task predictions. Still, existing methods suffer from two limitations: 1) The differentiated information between textual characters and Pinyin pronunciation are underexplored. 2) The task predictions are performed with the over-emphasises on either the textual or phonetic sequence, ignoring the balanced modeling and adaptive fusion. In this work, we proposed a method to alleviate the issues above. For the first issue, a Coupled Attention Module (CAM) is proposed where a couple of attention functions capture the associated and differentiated text-phonetics information simultaneously. For the second issue, an adaptive fusion is designed to derive the phonetics-aware textual representations and text-aware phonetic representations for task predictions. Experiments on three public benchmarks demonstrate the effectiveness of our proposed method.

Aspect sentiment quad prediction (ASQP) aims to predict the quad sentiment elements for a given sentence, which is a critical task in the field of aspect-based sentiment analysis. However, the data imbalance issue has not received sufficient attention in ASQP task. In this paper, we divide the issue into two-folds, quad-pattern imbalance and aspect-category imbalance, and propose an Adaptive Data Augmentation (ADA) framework to tackle the imbalance issue. Specifically, a data augmentation process with a condition function adaptively enhances the tail quad patterns and aspect categories, alleviating the data imbalance in ASQP. Following previous studies, we also further explore the generative framework for extracting complete quads by introducing the category prior knowledge and syntax-guided decoding target. Experimental1results demonstrate that data augmentation for imbalance in ASQP task can improve the performance, and the proposed ADA method is superior to naive data oversampling.

MultiModal Sentence Summarization (MMSS) aims to generate a brief summary based on the given source sentence and its associated image. Previous studies on MMSS have achieved success by either selecting the task-relevant visual information or filtering out the task-irrelevant visual information to help the textual modality to generate the summary. However, enhancing from a single perspective usually introduces over-preservation or over-compression problems. To tackle these issues, we resort to Information Bottleneck (IB), which seeks to find a maximally compressed mapping of the input information that preserves as much information about the target as possible. Specifically, we propose a novel method, T(3), which adopts IB to balance the Trade-off between Task-relevant and Task-irrelevant visual information through the variational inference framework. In this way, the task-irrelevant visual information is compressed to the utmost while the task-relevant visual information is maximally retained. With the holistic perspective, the generated summary could maintain as many key elements as possible while discarding the unnecessary ones as far as possible. Extensive experiments on the representative MMSS dataset demonstrate the superiority of our proposed method. Our code is available at https://github.com/YuanMinghuan/T3.

Low-resource Complex Named Entity Recognition aims to detect entities with the form of any linguistic constituent under scenarios with limited manually annotated data. Existing studies augment the text through the substitution of same type entities or language modeling, but suffer from the lower quality and the limited entity context patterns within low-resource corpora. In this paper, we propose a novel data augmentation method E2DA from both exogenous and endogenous perspectives. As for exogenous augmentation, we treat the limited manually annotated data as anchors, and leverage the powerful instruction-following capabilities of Large Language Models (LLMs) to expand the anchors by generating data that are highly dissimilar from the original anchor texts in terms of entity mentions and contexts. As regards the endogenous augmentation, we explore diverse semantic directions in the implicit feature space of the original and expanded anchors for effective data augmentation. Our complementary augmentation method from two perspectives not only continuously expands the global text-level space, but also fully explores the local semantic space for more diverse data augmentation. Extensive experiments on 10 diverse datasets across various low-resource settings demonstrate that the proposed method excels significantly over prior state-of-the-art data augmentation methods.

This paper studies the multimodal named entity recognition (MNER) and multimodal relation extraction (MRE), which are important for content analysis and various applications. The core of MNER and MRE lies in incorporating evident visual information to enhance textual semantics, where two issues inherently demand investigations. The first issue is modality-noise, where the task-irrelevant information in each modality may be noises misleading the task prediction. The second issue is modality-gap, where representations from different modalities are inconsistent, preventing from building the semantic alignment between the text and image. To address these issues, we propose a novel method for MNER and MRE byMultiModal representation learning withInformationBottleneck (MMIB). For the first issue, a refinement-regularizer probes the information-bottleneck principle to balance the predictive evidence and noisy information, yielding expressive representations for prediction. For the second issue, an alignment-regularizer is proposed, where a mutual information-based item works in a contrastive manner to regularize the consistent text-image representations. To our best knowledge, we are the first to explore variational IB estimation for MNER and MRE. Experiments show that MMIB achieves the state-of-the-art performances on three public benchmarks.

Event extraction (EE) is a crucial research task for promptly apprehending event information from massive textual data. With the rapid development of deep learning, EE based on deep learning technology has become a research hotspot. Numerous methods, datasets, and evaluation metrics have been proposed in the literature, raising the need for a comprehensive and updated survey. This article fills the research gap by reviewing the state-of-the-art approaches, especially focusing on the general domain EE based on deep learning models. We introduce a new literature classification of current general domain EE research according to the task definition. Afterward, we summarize the paradigm and models of EE approaches, and then discuss each of them in detail. As an important aspect, we summarize the benchmarks that support tests of predictions and evaluation metrics. A comprehensive comparison among different approaches is also provided in this survey. Finally, we conclude by summarizing future research directions facing the research area.

Darknet markets provide a large platform for trading illicit goods and services due to their anonymity. Learning an invariant representation of each user based on their posts on different markets makes it easy to aggregate user information across different platforms, which helps identify anonymous users. Traditional user representation methods mainly rely on modeling the text information of posts and cannot capture the temporal content and the forum interaction of posts. While recent works mainly use CNN to model the text information of posts, failing to effectively model posts whose length changes frequently in an episode. To address the above problems, we propose a model named URM4DMU(User Representation Model for Darknet Markets Users) which mainly improves the post representation by augmenting convolutional operators and self-attention with an adaptive gate mechanism. It performs much better when combined with the temporal content and the forum interaction of posts. We demonstrate the effectiveness of URM4DMU on four darknet markets. The average improvements on MRR value and Recall@10 are 22.5% and 25.5% over the state-of-the-art method respectively.

Event Causality Identification (ECI), which aims to detect whether a causality relation exists between two given textual events, is an important task for event causality understanding. However, the ECI task ignores crucial event structure and cause-effect causality component information, making it struggle for downstream applications. In this paper, we introduce a novel task, namely Event Causality Extraction (ECE), aiming to extract the cause-effect event causality pairs with their structured event information from plain texts. The ECE task is more challenging since each event can contain multiple event arguments, posing fine-grained correlations between events to decide the cause-effect event pair. Hence, we propose a method with a dual grid tagging scheme to capture the intra- and inter-event argument correlations for ECE. Further, we devise a event type-enhanced model architecture to realize the dual grid tagging scheme. Experiments demonstrate the effectiveness of our method, and extensive analyses point out several future directions for ECE.

Document-level Event Extraction (DEE) is particularly tricky due to the two challenges it poses: scattering-arguments and multi-events. The first challenge means that arguments of one event record could reside in different sentences in the document, while the second one reflects that one document may simultaneously contain multiple such event records. Motivated by humans’ reading cognitive to extract information of interests, in this paper, we propose a method called HRE (Human Reading inspired Extractor for Document Events), where DEE is decomposed into these two iterative stages, rough reading and elaborate reading. Specifically, the first stage browses the document to detect the occurrence of events, and the second stage serves to extract specific event arguments. For each concrete event role, elaborate reading hierarchically works from sentences to characters to locate arguments across sentences, thus the scattering-arguments problem is tackled. Meanwhile, rough reading is explored in a multi-round manner to discover undetected events, thus the multi-events problem is handled. Experiment results show the superiority of HRE over prior competitive methods.

In few-shot relational triple extraction (FS-RTE), one seeks to extract relational triples from plain texts by utilizing only few annotated samples. Recent work first extracts all entities and then classifies their relations. Such an entity-then-relation paradigm ignores the entity discrepancy between relations. To address it, we propose a novel task decomposition strategy, Relation-then-Entity, for FS-RTE. It first detects relations occurred in a sentence and then extracts the corresponding head/tail entities of the detected relations. To instantiate this strategy, we further propose a model, RelATE, which builds a dual-level attention to aggregate relation-relevant information to detect the relation occurrence and utilizes the annotated samples of the detected relations to extract the corresponding head/tail entities. Experimental results show that our model outperforms previous work by an absolute gain (18.98%, 28.85% in F1 in two few-shot settings).

Event detection (ED) is a pivotal task for information retrieval, which aims at identifying event triggers and classifying them into pre-defined event types. In real-world applications, events are usually annotated with numerous fine-grained types, which often arises long-tail type nature and co-occurrence event nature. Existing studies explore the event correlations without full utilization, which may limit the capability of event detection. This paper simultaneously incorporates both the type-level and instance-level event correlations, and proposes a novel framework, termed as CorED. Specifically, we devise an adaptive graph-based type encoder to capture instance-level correlations, learning type representations not only from their training data but also from their relevant types, thus leading to more informative type representations especially for the low-resource types. Besides, we devise an instance interactive decoder to capture instance-level correlations, which predicts event instance types conditioned on the contextual typed event instances, leveraging co-occurrence events as remarkable evidence in prediction. We conduct experiments on two public benchmarks, MAVEN and ACE-2005 dataset. Empirical results demonstrate the unity of both type-level and instance-level correlations, and the model achieves effectiveness performance on both benchmarks.

To obtain the specific location information of personnel in the underground mine, this paper describes a real-time locating system (RTLS) based on ultra-wideband (UWB) technology. The system consists of wearable active tags, main and sub anchors, servers, and clients. Firstly, alternative double-sided two-way ranging (ADS-TWR) combined with digital filter and linear fitting is used to pre-process the noise and bias of the ranging results, obtaining the anchor-tag distance with an error less than 10cm. In order to resolve the problem of position outliers caused by random factors in the mine, the position of the moving target is estimated by the extended Kalman filter (EKF) initially. Then the mean filter (MF) will perform a secondary optimization of the results. Experimental results show that the positioning accuracy of the system reaches 0.06m in the line-of-sight environment. The positioning accuracy is improved effectively and the system shows a strong anti-noise performance.