Large Language Models have shown growing ability to generate fluent and coherent texts that are highly similar to the writing style of humans. Current detectors for Machine-Generated Text (MGT) perform well when they are trained and tested in the same domain but generalize poorly to unseen domains, due to domain shift between data from different sources. In this work, we propose MGT-Prism , an MGT detection method from the perspective of the frequency domain for better domain generalization. Our key insight stems from analyzing text representations in the frequency domain, where we observe consistent spectral patterns across diverse domains, while significant discrepancies in magnitude emerge between MGT and human-written texts (HWTs). The observation initiates the design of a low frequency domain filtering module for filtering out the document-level features that are sensitive to domain shift, and a dynamic spectrum alignment strategy to extract the task-specific and domain-invariant features for improving the detector's performance in domain generalization. Extensive experiments demonstrate that MGT-Prism outperforms state‑of‑the‑art baselines by an average of 0.90% in accuracy and 0.92% in F1 score on 11 test datasets across three domain‑generalization scenarios.

Chengzhengxu Li, Xiaoming Liu, Zhaohan Zhang, Shengchao Liu, Guoxin Ma, Yu Lan, Cong Wang, Chao Shen. Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2026.

Zhaohan Zhang, Chengzhengxu Li, Xiaoming Liu, Chao Shen, Ziquan Liu, Ioannis Patras. Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2026.

Assessing the reliability of Large Language Models (LLMs) by confidence elicitation is a prominent approach to AI safety in highstakes applications, such as healthcare and finance.Existing methods either require expensive computational overhead or suffer from poor calibration, making them impractical and unreliable for real-world deployment.In this work, we propose GrACE, a Generative Approach to Confidence Elicitation that enables scalable and reliable confidence elicitation for LLMs.GrACE adopts a novel mechanism in which the model expresses confidence by the similarity between the last hidden state and the embedding of a special token appended to the vocabulary, in real-time.We fine-tune the model for calibrating the confidence with targets associated with accuracy.Extensive experiments show that the confidence produced by GrACE achieves the best discriminative capacity and calibration on open-ended generation tasks without resorting to additional sampling or an auxiliary model.Moreover, we propose two confidence-based strategies for test-time scaling with GrACE, which not only improve the accuracy of the final decision but also significantly reduce the number of required samples, highlighting its potential as a practical solution for deploying LLMs with reliable, on-the-fly confidence estimation.The code is available at: https://github.com/petezone/Grace.

Machine-generated Text (MGT) detection is crucial for regulating and attributing online texts.While the existing MGT detectors achieve strong performance, they remain vulnerable to simple perturbations and adversarial attacks.To build an effective defense against malicious perturbations, we view MGT detection from a threat modeling perspective, that is, analyzing the model's vulnerability from an adversary's point of view and exploring effective mitigations.To this end, we introduce an adversarial framework for training a robust MGT detector, named GREedy Adversary PromoTed DefendER (GREATER).The GREATER consists of two key components: an adversary GREATER-A and a detector GREATER-D.The GREATER-D learns to defend against the adversarial attack from GREATER-A and generalizes the defense to other attacks.GREATER-A identifies and perturbs the critical tokens in embedding space, along with greedy search and pruning to generate stealthy and disruptive adversarial examples.Besides, we update the GREATER-A and GREATER-D synchronously, encouraging the GREATER-D to generalize its defense to different attacks and varying attack intensities.Our experimental results across 10 text perturbation strategies and 6 adversarial attacks show that our GREATER-D reduces the Attack Success Rate (ASR) by 0.67% compared with SOTA defense methods while our GREATER-A is demonstrated to be more effective and efficient than SOTA attack approaches.Codes and dataset are available in https:// github.com/Liyuuuu111/GREATER.

The misuse of large language models (LLMs) poses potential risks, motivating the development of machine-generated text (MGT) detection. Existing literature primarily concentrates on binary, document-level detection, thereby neglecting texts that are composed jointly by human and LLM contributions. Hence, this paper explores the possibility of fine-grained MGT detection under human-AI coauthoring.We suggest fine-grained detectors can pave pathways toward coauthored text detection with a numeric AI ratio.Specifically, we propose a dataset, HACo-Det, which produces human-AI coauthored texts via an automatic pipeline with word-level attribution labels. We retrofit seven prevailing document-level detectors to generalize them to word-level detection.Then we evaluate these detectors on HACo-Det on both word- and sentence-level detection tasks.Empirical results show that metric-based methods struggle to conduct fine-grained detection with a 0.462 average F1 score, while finetuned models show superior performance and better generalization across domains. However, we argue that fine-grained co-authored text detection is far from solved.We further analyze factors influencing performance, e.g., context window, and highlight the limitations of current methods, pointing to potential avenues for improvement.

Large Language Models (LLMs) have been found to memorize and recite some of the textual sequences from their training set verbatim, raising broad concerns about privacy and copyright issues. This Textual Sequence Memorization (TSM) phenomenon leads to a high demand to regulate LLM output to prevent generating certain memorized text that a user wants to be forgotten. However, our empirical study reveals that existing methods for TSM erasure fail to unlearn large numbers of memorized samples without substantially jeopardizing the model utility. To achieve a better trade-off between the effectiveness of TSM erasure and model utility in LLMs, our paper proposes a new method, named Entropy Maximization with Selective Optimization (EMSO), where the model parameters are updated sparsely based on novel optimization and selection criteria, in a manner that does not require additional models or data other than that in the forget set. More specifically, we propose an entropy-based loss that is shown to lead to more stable optimization and better preserves model utility than existing methods. In addition, we propose a contrastive gradient metric that takes both the gradient magnitude and direction into consideration, so as to localize model parameters to update in a sparse model updating scehme. Extensive experiments across three model scales demonstrate that our method excels in handling large-scale forgetting requests while preserving model ability in language generation and understanding.

The detection and segmentation of thyroid nodules in ultrasound images pose considerable challenges due to significant variability in the shape and size of nodules, as well as the influence of artifacts and background noise. To mitigate these challenges, we introduce Bidirectional Attention YOLO (BDA-YOLO), an advanced YOLOv11 model that integrates two innovative feature fusion modules: the Dynamic Self-Attention Module (DSAM) and the Bidirectional Fusion Module (BDFM). The DSAM processes input feature maps through convolution and flattening prior to the application of self-attention mechanisms, facilitating a more precise integration of multi-level feature information. Meanwhile, BDFM enhances feature interactions across different layers, thereby improving the network’s capacity to capture intricate characteristics of nodules. Experimental evaluations conducted on the TN3K and DDTI datasets indicate that our proposed method exceeds existing state-of-the-art models, including RT-DETRv1 and RT-DETRv2, achieving superior Box Average Precision (AP) and Mask AP metrics, alongside notable enhancements in precision and recall. These findings highlight the efficacy of our feature fusion approach in improving both the detection and segmentation outcomes for the localization of thyroid nodules on ultrasound imaging.

Accurate segmentation of brain tumors, especially gliomas, is essential for medical diagnosis, treatment planning, and monitoring of disease progression. Traditional manual segmentation methods are time-consuming and prone to variability, while deep learning-based approaches, such as convolutional neural networks (CNNs), have significantly improved performance. However, CNNs often struggle to capture global context due to their limited receptive fields. To address these challenges, we propose SwinFusion-XL, a novel segmentation model based on the Swin UNETR architecture that integrates local and global features for enhanced brain tumor segmentation. SwinFusion-XL leverages a Superficial Mamba Fusion Module (SMFM) for the effective integration of multi-scale features, combining both channel and spatial attention mechanisms to enhance feature representations. Additionally, a Cross xLSTM Module (CxLSTM) is introduced to model cross-feature dependencies and capture long-range spatial, channel, and volumetric relationships. Extensive experiments on the BraTS2020 and BraTS2021 brain tumor datasets demonstrate that our model consistently outperforms state-of-the-art methods, achieving significant improvements in segmentation performance.

Shengchao Liu, Xiaoming Liu, Yichen Wang, Zehua Cheng, Chengzhengxu Li, Zhaohan Zhang, Yu Lan, Chao Shen. Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2024.

Recent advances in prompt optimization have notably enhanced the performance of pre-trained language models (PLMs) on downstream tasks. However, the potential of optimized prompts on domain generalization has been under-explored. To explore the nature of prompt generalization on unknown domains, we conduct pilot experiments and find that (i) Prompts gaining more attention weight from PLMs’ deep layers are more generalizable and (ii) Prompts with more stable attention distributions in PLMs’ deep layers are more generalizable. Thus, we offer a fresh objective towards domain-generalizable prompts optimization named ''Concentration'', which represents the ''lookback'' attention from the current decoding token to the prompt tokens, to increase the attention strength on prompts and reduce the fluctuation of attention distribution. We adapt this new objective to popular soft prompt and hard prompt optimization methods, respectively. Extensive experiments demonstrate that our idea improves comparison prompt optimization methods by 1.42% for soft prompt generalization and 2.16% for hard prompt generalization in accuracy on the multi-source domain generalization setting, while maintaining satisfying in-domain performance. The promising results validate the effectiveness of our proposed prompt optimization objective and provide key insights into domain-generalizable prompts.

Machine-Generated Text (MGT) detection, a task that discriminates MGT from Human-Written Text (HWT), plays a crucial role in preventing misuse of text generative models, which excel in mimicking human writing style recently.The latest proposed detectors usually take coarse text sequences as input and finetune pre-trained models with standard crossentropy loss.However, these methods fail to consider the linguistic structure of texts.Moreover, they lack the ability to handle the lowresource problem, which could often happen in practice considering the enormous amount of textual data online.In this paper, we present a coherence-based contrastive learning model named COCO to detect the possible MGT under the low-resource scenario.To exploit the linguistic feature, we encode coherence information in the form of graph into the text representation.To tackle the challenges of low data resources, we employ a contrastive learning framework and propose an improved contrastive loss for preventing performance degradation brought by simple samples.The experiment results on two public datasets and two self-constructed datasets prove our approach outperforms the state-of-the-art methods significantly.Also, we surprisingly find that MGTs originated from up-to-date language models could be easier to detect than these from previous models, in our experiments.And we propose some preliminary explanations for this counter-intuitive phenomena.All the codes and datasets are open-sourced.1

Accurate traffic anomaly prediction offers an opportunity to save the wounded at the right location in time. However, the complex process of traffic anomaly is affected by both various static factors and dynamic interactions. The recent evolving representation learning provides a new possibility to understand this complicated process, but with challenges of imbalanced data distribution and heterogeneity of features. To tackle these problems, this paper proposes a spatio-temporal evolution model namedSNIPERfor learning intricate feature interactions to predict traffic anomalies. Specifically, we design spatio-temporal encoders to transform spatio-temporal information into vector space indicating their natural relationship. Then, we propose a temporally dynamical evolving embedding method to pay more attention to rare traffic anomalies and develop an effective attention-based multiple graph convolutional network to formulate the spatially mutual influence from three different perspectives. The FC-LSTM is adopted to aggregate the heterogeneous features considering the spatio-temporal influences. Finally, a loss function is designed to overcome the ’over-smoothing’ and solve the imbalanced data problem. Extensive experiments show that SNIPER averagely outperforms state-of-the-arts by 3.9%, 0.9%, 1.9% and 1.6% on Chicago datasets, and 2.4%, 0.6%, 2.6% and 1.3% on New York City datasets in metrics of AUC-PR, AUC-ROC, F1 score, and accuracy, respectively.

Knowledge graph (KG), integrating complex information and containing rich semantics, is widely considered as side information to enhance the recommendation systems. However, most of the existing KG-based methods concentrate on encoding the structural information in the graph, without utilizing the collaborative signals in user-item interaction data, which are important for understanding user preferences. Therefore, the representations learned by these models are insufficient for representing semantic information of users and items in the recommendation environment. The combination of both kinds of data provides a good chance to solve this problem, but it faces the following challenges: i) the inner correlations in user-item interaction data are difficult to capture from one side of the user or item; ii) capturing the knowledge associations on the whole KG would introduce noises and variously influence the recommendation results; iii) the semantic gap between both kinds of data is hard to alleviate.