The effective detection and governance of Large Language Model (LLM) generated content has become increasingly critical due to the growing risk of misuse. Despite the impressive performance of existing detectors, their reliability and potential in multilingual, real-world scenarios remain largely underexplored.In this study, we introduce DetectRL-X, a comprehensive multilingual benchmark designed to evaluate advanced detectors across 8 dimensions. The benchmark encompasses 8 languages commonly used in commercial contexts and collects human-written texts from 6 domains highly susceptible to LLM misuse. To better aligned with real-world applications, We create LLM-generated texts using 4 popular commercial LLMs, and include typical AI-assisted writing operations such as polishing, expanding, and condensing to capture authentic usage patterns. Furthermore, we develop a multilingual framework for paraphrasing and perturbation attacks to simulate diverse human modifications and writing noise, enabling stress testing of detectors across languages.Experimental results on DetectRL-X reveal the strengths and limitations of current state-of-the-art detectors when applied to diverse linguistic resources. We further analyze how domains, generators, attack strategies, text length, and refinement operations influence performance in different languages, underscoring DetectRL-X as an effective benchmark for strengthening multilingual and language-specific detectors.

Jinghan Yu, Junhao Xiao, Chenyu Zhu, Jiaming Li, Jia Li, HanMing Deng, Xirui Wang, Guoli Jia, Jianjun Li, Xiang Bai, Bowen Zhou, Zhiyuan Ma. Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2026.

3D shape segmentation is an essential task in computer graphics and is widely used in many applications. It plays a critical role in understanding the structure and semantics of 3D models. Traditional approaches to 3D shape segmentation primarily rely on geometric features to partition models into meaningful components. However, these methods often struggle when the geometric characteristics of different parts are similar, resulting in ambiguous segmentation outcomes. To address this fundamental limitation, we introduce a novel text-driven multi-modal framework that systematically integrates textual semantics with geometric analysis for enhanced 3D shape segmentation. Our approach leverages a pre-trained language model with prefix tuning to bridge the semantic granularity gap between part-level annotations and face-level segmentation, while a specialized mesh self-attention module captures contextual relationships among neighboring faces. We design an attention-based text-driven integration mechanism that dynamically weights multimodal features, complemented by a Laplace-Adaptive Attention Module (LAAM) that better handles the distributions of geometric features. Through contrastive learning, we align textual and geometric representations in a shared semantic space, enabling effective disambiguation of geometrically similar but semantically distinct parts. We also contribute the Fine-grained HumanBody benchmark for comprehensive evaluation. Extensive experiments on Princeton Segmentation Benchmark, COSEG, ShapeNetCore, and our proposed benchmark demonstrate that our method significantly outperforms existing approaches, achieving superior segmentation accuracy while effectively resolving geometric ambiguities through semantic understanding.

Bo Zeng, Chenyang Lyu, Sinuo Liu, Mingyan Zeng, Minghao Wu, Xuanfan Ni, Tianqi Shi, Yu Zhao, Yefeng Liu, Chenyu Zhu, Ruizhe Li, Jiahui Geng, Qing Li, Yu Tong, Longyue Wang, Weihua Luo, Kaifu Zhang. Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2025.

Fetal congenital heart disease (FCHD) is a major cause of perinatal mortality, yet accurate prenatal echocardiographic screening remains challenged by complex fetal anatomy, variable fetal posture, and an abundance of non-diagnostic frames. Although deep learning has accelerated automated echocardiography, dedicated solutions for fetal imaging are scarce. We compiled 388 expertly annotated fetal 2-D echocardiogram videos (284 normal, 104 FCHD) and introduce an end-to-end framework that: (i) isolates diagnostic frames with a Monte Carlo Keyframe Selector, (ii) pinpoints salient anatomy with an Evidence Region Extractor, and (iii) integrates local and global cues via a Cross Retrieval Module. Extensive evaluations demonstrate substantial gains in diagnostic robustness and accuracy, underscoring the framework's potential to advance prenatal cardiac screening.

Object detection models are vulnerable to backdoor attacks, where attackers poison a small subset of training samples by embedding a predefined trigger to manipulate prediction. Detecting poisoned samples (i.e., those containing triggers) at test time can prevent backdoor activation. However, unlike image classification tasks, the unique characteristics of object detection—particularly its output of numerous objects—pose fresh challenges for backdoor detection. The complex attack effects (e.g., "ghost" object emergence or "vanishing" object) further render current defenses fundamentally inadequate. To this end, we design TRAnsformation Consistency Evaluation (TRACE), a brand-new method for detecting poisoned samples at test time in object detection. Our journey begins with two intriguing observations: 1) poisoned samples exhibit significantly more consistent detection results than clean ones across varied backgrounds. 2) clean samples show higher detection consistency when introduced to different focal information. Based on these phenomena, Trace applies foreground and background transformations to each test sample, then assesses transformation consistency by calculating the variance in objects confidences. Trace achieves black-box, universal backdoor detection, with extensive experiments showing a 30% improvement in AUROC over state-of-the-art defenses and resistance to adaptive attacks.

Embodied AI represents systems where AI is integrated into physical entities. Multimodal Large Language Model (LLM), which exhibits powerful language understanding abilities, has been extensively employed in embodied AI by facilitating sophisticated task planning. However, a critical safety issue remains overlooked: could these embodied LLMs perpetrate harmful behaviors? In response, we introduce BadRobot, the first attack paradigm designed to jailbreak robotic manipulation, making embodied LLMs violate safety and ethical constraints through typical voice-based user-system interactions. Specifically, three vulnerabilities are exploited to achieve this type of attack: (i) manipulation of LLMs within robotic systems, (ii) misalignment between linguistic outputs and physical actions, and (iii) unintentional hazardous behaviors caused by world knowledge's flaws. Furthermore, we construct a benchmark of various malicious physical action queries to evaluate BadRobot's attack performance. Based on this benchmark, extensive experiments against existing prominent embodied LLM frameworks (e.g., Voxposer, Code as Policies, and ProgPrompt) demonstrate the effectiveness of our BadRobot. We emphasize that addressing this emerging vulnerability is crucial for the secure deployment of LLMs in robotics. Warning: This paper contains harmful AI-generated language and aggressive actions.

Despite their outstanding performance in numerous applications, large language models (LLMs) remain prone to hallucinations, generating content inconsistent with their pretraining corpora. Currently, almost all contrastive decoding approaches alleviate hallucinations by introducing a model susceptible to hallucinations and appropriately widening the contrastive logits gap between hallucinatory tokens and target tokens. However, although existing contrastive decoding methods mitigate hallucinations, they lack enough confidence in the factual accuracy of the generated content. In this work, we propose Multi-Model Contrastive Decoding (MCD), which integrates a pretrained language model with an evil model and a truthful model for contrastive decoding. Intuitively, a token is assigned a high probability only when deemed potentially hallucinatory by the evil model while being considered factual by the truthful model. This decoding strategy significantly enhances the model’s confidence in its generated responses and reduces potential hallucinations. Furthermore, we introduce a dynamic hallucination detection mechanism that facilitates token-by-token identification of hallucinations during generation and a tree-based revision mechanism to diminish hallucinations further. Extensive experimental evaluations demonstrate that our MCD strategy effectively reduces hallucinations in LLMs and outperforms state-of-the-art methods across various benchmarks.

In this paper, the modeling and mitigation method of torque ripple are investigated for permanent-magnet synchronous motor (PMSM). A new torque ripple model is proposed for three-phase PMSM based on general airgap field modulation theory. The mapping relationship between the torque order and the current order has been revealed. The qualitative and quantitative relationship between the stator harmonic current and torque ripple is established directly without other intermediate variables. The phase and amplitude of the torque component are found to be proportional to the phase and amplitude of the injected harmonic current. Therefore, only a few constant coefficients are required by finite element analysis (FEA) simulation. With the proposed torque model, directional mitigation of torque ripple can be realized by harmonic current injection method with low computational burden by FEA. Repetitive control is employed for harmonic current injection. FEA simulation and experimental results at different operating conditions verify the validity of the proposed torque model and ripple mitigation method.

Yawning detection is actively used in multimedia applications such as driver fatigue assessment and status monitoring. However, the accuracy and robustness of existing yawning detectors are limited due to variations in environments (especially lights), facial expressions, and confusion behaviours (e.g., talking and eating). This paper introduces a transformer-based method, YawnNet, for accurate yawning detection by leveraging spatial-temporal encoding and local cues. In particular, YawnNet contains a data processing stage with temporal downsampling and cube embedding on the input sequence. Moreover, it includes a Swin-Transformer block that operates on fine-grained patches to uncover short-range local cues. Through comprehensive experiments, we demonstrate the advantages of YawnNet: (1) significantly higher accuracy than the state-of-the-art Dense-LSTM (precision and recall increased by 2.3% and 4.2%, respectively) on the FatigueView dataset, (2) close to real-time (30 FPS on RTX 3090), and (3) a marked improvement in robustness on confusion behaviours, invariance (resolution and orientation) and complex scenarios (occlusion, over- and underexpose).

Emerging viral infections seriously threaten human health globally. Several challenges exist in identifying effective compounds against viral infections: (1) at the initial stage of a new virus outbreak, little information, except for its genome information, may be available; (2) although the identified compounds may be effective, they may be toxic in vivo and (3) cytokine release syndrome (CRS) triggered by viral infections is the primary cause of mortality. Currently, an integrative tool that takes all those aspects into consideration for identifying effective compounds to prevent viral infections is absent. In this study, we developed iDMer, as an integrative and mechanism-driven response system for addressing these challenges during the sudden virus outbreaks. iDMer comprises three mechanism-driven compound identification modules, that is, a virus-host interaction-oriented module, an autophagy-oriented module and a CRS-oriented module. As a one-stop integrative platform, iDMer incorporates compound toxicity evaluation and compound combination identification for virus treatment with clear mechanisms. iDMer was successfully tested on five viruses, including the current severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2). Our results indicated that, for all five tested viruses, compounds that were reported in the literature or experimentally validated for virus treatment were enriched at the top, demonstrating the generalized effectiveness of iDMer. Finally, we demonstrated that combinations of the individual modules successfully identified combinations of compounds effective for virus intervention with clear mechanisms.