Jie Cao

Huaibo Huang, Xiaoqiang Zhou, Jie Cao et al.

Vision transformer has achieved impressive performance for many vision tasks. However, it may suffer from high redundancy in capturing local features for shallow layers. Local self-attention or early-stage convolutions are thus utilized, which sacrifice the capacity to capture long-range dependency. A challenge then arises: can we access efficient and effective global context modeling at the early stages of a neural network? To address this issue, we draw inspiration from the design of superpixels, which reduces the number of image primitives in subsequent processing, and introduce super tokens into vision transformer. Super tokens attempt to provide a semantically meaningful tessellation of visual content, thus reducing the token number in self-attention as well as preserving global modeling. Specifically, we propose a simple yet strong super token attention (STA) mechanism with three steps: the first samples super tokens from visual tokens via sparse association learning, the second performs self-attention on super tokens, and the last maps them back to the original token space. STA decomposes vanilla global attention into multiplications of a sparse association map and a low-dimensional attention, leading to high efficiency in capturing global dependencies. Based on STA, we develop a hierarchical vision transformer. Extensive experiments demonstrate its strong performance on various vision tasks. In particular, without any extra training data or label, it achieves 86.4% top-1 accuracy on ImageNet-1K with less than 100M parameters. It also achieves 53.9 box AP and 46.8 mask AP on the COCO detection task, and 51.9 mIOU on the ADE20K semantic segmentation task. Code is released at https://github.com/hhb072/STViT.

Jijie Wu, Dongliang Chang, Aneeshan Sain et al.

The main challenge for fine-grained few-shot image classification is to learn feature representations with higher inter-class and lower intra-class variations, with a mere few labelled samples. Conventional few-shot learning methods however cannot be naively adopted for this fine-grained setting -- a quick pilot study reveals that they in fact push for the opposite (i.e., lower inter-class variations and higher intra-class variations). To alleviate this problem, prior works predominately use a support set to reconstruct the query image and then utilize metric learning to determine its category. Upon careful inspection, we further reveal that such unidirectional reconstruction methods only help to increase inter-class variations and are not effective in tackling intra-class variations. In this paper, we for the first time introduce a bi-reconstruction mechanism that can simultaneously accommodate for inter-class and intra-class variations. In addition to using the support set to reconstruct the query set for increasing inter-class variations, we further use the query set to reconstruct the support set for reducing intra-class variations. This design effectively helps the model to explore more subtle and discriminative features which is key for the fine-grained problem in hand. Furthermore, we also construct a self-reconstruction module to work alongside the bi-directional module to make the features even more discriminative. Experimental results on three widely used fine-grained image classification datasets consistently show considerable improvements compared with other methods. Codes are available at: https://github.com/PRIS-CV/Bi-FRN.

Jin Liu, Huaibo Huang, Jie Cao et al.

Diffusion-based text-to-image generation models have significantly advanced the field of art content synthesis. However, current portrait stylization methods generally require either model fine-tuning based on examples or the employment of DDIM Inversion to revert images to noise space, both of which substantially decelerate the image generation process. To overcome these limitations, this paper presents an inversion-free portrait stylization framework based on diffusion models that accomplishes content and style feature fusion in merely four sampling steps. We observed that Latent Consistency Models employing consistency distillation can effectively extract representative Consistency Features from noisy images. To blend the Consistency Features extracted from both content and style images, we introduce a Style Enhancement Attention Control technique that meticulously merges content and style features within the attention space of the target image. Moreover, we propose a feature merging strategy to amalgamate redundant features in Consistency Features, thereby reducing the computational load of attention control. Extensive experiments have validated the effectiveness of our proposed framework in enhancing stylization efficiency and fidelity. The code is available at \url{https://github.com/liujin112/ZePo}.

Changyu Chen, Xiting Wang, Yiqiao Jin et al. · gatech

In reinforcement learning (RL), there are two major settings for interacting with the environment: online and offline. Online methods explore the environment at significant time cost, and offline methods efficiently obtain reward signals by sacrificing exploration capability. We propose semi-offline RL, a novel paradigm that smoothly transits from offline to online settings, balances exploration capability and training cost, and provides a theoretical foundation for comparing different RL settings. Based on the semi-offline formulation, we present the RL setting that is optimal in terms of optimization cost, asymptotic error, and overfitting error bound. Extensive experiments show that our semi-offline approach is efficient and yields comparable or often better performance compared with state-of-the-art methods.

Jie Cao, Dian Jiao, Qiang Yan et al.

Query-focused summarization (QFS) aims to produce summaries that answer particular questions of interest, enabling greater user control and personalization. With the advent of large language models (LLMs), shows their impressive capability of textual understanding through large-scale pretraining, which implies the great potential of extractive snippet generation. In this paper, we systematically investigated two indispensable characteristics that the LLMs-based QFS models should be harnessed, Lengthy Document Summarization and Efficiently Fine-grained Query-LLM Alignment, respectively. Correspondingly, we propose two modules called Query-aware HyperExpert and Query-focused Infini-attention to access the aforementioned characteristics. These innovations pave the way for broader application and accessibility in the field of QFS technology. Extensive experiments conducted on existing QFS benchmarks indicate the effectiveness and generalizability of the proposed approach. Our code is publicly available at https://github.com/DCDmllm/IDEAL_Summary.

Jon Z. Cai, Brendan King, Margaret Perkoff et al.

In this paper, we introduce Dependency Dialogue Acts (DDA), a novel framework for capturing the structure of speaker-intentions in multi-party dialogues. DDA combines and adapts features from existing dialogue annotation frameworks, and emphasizes the multi-relational response structure of dialogues in addition to the dialogue acts and rhetorical relations. It represents the functional, discourse, and response structure in multi-party multi-threaded conversations. A few key features distinguish DDA from existing dialogue annotation frameworks such as SWBD-DAMSL and the ISO 24617-2 standard. First, DDA prioritizes the relational structure of the dialogue units and the dialog context, annotating both dialog acts and rhetorical relations as response relations to particular utterances. Second, DDA embraces overloading in dialogues, encouraging annotators to specify multiple response relations and dialog acts for each dialog unit. Lastly, DDA places an emphasis on adequately capturing how a speaker is using the full dialog context to plan and organize their speech. With these features, DDA is highly expressive and recall-oriented with regard to conversation dynamics between multiple speakers. In what follows, we present the DDA annotation framework and case studies annotating DDA structures in multi-party, multi-threaded conversations.

Chun Bao, Jie Cao, Yaqian Ning et al.

Infrared small target detection based on deep learning offers unique advantages in separating small targets from complex and dynamic backgrounds. However, the features of infrared small targets gradually weaken as the depth of convolutional neural network (CNN) increases. To address this issue, we propose a novel method for detecting infrared small targets called improved dense nested attention network (IDNANet), which is based on the transformer architecture. We preserve the dense nested structure of dense nested attention network (DNANet) and introduce the Swin-transformer during feature extraction stage to enhance the continuity of features. Furthermore, we integrate the ACmix attention structure into the dense nested structure to enhance the features of intermediate layers. Additionally, we design a weighted dice binary cross-entropy (WD-BCE) loss function to mitigate the negative impact of foreground-background imbalance in the samples. Moreover, we develop a dataset specifically for infrared small targets, called BIT-SIRST. The dataset comprises a significant amount of real-world targets and manually annotated labels, as well as synthetic data and corresponding labels. We have evaluated the effectiveness of our method through experiments conducted on public datasets. In comparison to other state-of-the-art methods, our approach outperforms in terms of probability of detection ($P_d$), false-alarm rate ($F_a$), and mean intersection of union ($mIoU$). The $mIoU$ reaches 90.89\% on the NUDT-SIRST dataset and 79.72\% on the SIRST dataset. The BIT-SIRST dataset and codes are available openly at \href{https://github.com/EdwardBao1006/bit\_sirst}{\color[HTML]{B22222}{https://github.com/EdwardBao1006/bit\_sirst}}.

Jifan Yu, Zheyuan Zhang, Daniel Zhang-li et al.

Since the first instances of online education, where courses were uploaded to accessible and shared online platforms, this form of scaling the dissemination of human knowledge to reach a broader audience has sparked extensive discussion and widespread adoption. Recognizing that personalized learning still holds significant potential for improvement, new AI technologies have been continuously integrated into this learning format, resulting in a variety of educational AI applications such as educational recommendation and intelligent tutoring. The emergence of intelligence in large language models (LLMs) has allowed for these educational enhancements to be built upon a unified foundational model, enabling deeper integration. In this context, we propose MAIC (Massive AI-empowered Course), a new form of online education that leverages LLM-driven multi-agent systems to construct an AI-augmented classroom, balancing scalability with adaptivity. Beyond exploring the conceptual framework and technical innovations, we conduct preliminary experiments at Tsinghua University, one of China's leading universities. Drawing from over 100,000 learning records of more than 500 students, we obtain a series of valuable observations and initial analyses. This project will continue to evolve, ultimately aiming to establish a comprehensive open platform that supports and unifies research, technology, and applications in exploring the possibilities of online education in the era of large model AI. We envision this platform as a collaborative hub, bringing together educators, researchers, and innovators to collectively explore the future of AI-driven online education.

Tianwei Lin, Zhongwei Qiu, Jie Cao et al.

Medical vision-language models (VLMs) have rapidly advanced as general-purpose multimodal assistants, yet their deployment in 3D Computed Tomography (CT) analysis remains constrained by a persistent mismatch between optimization objectives and clinical rigor. Current Reinforcement Learning (RL) paradigms still rely on lexical proxy signals that induce ``\textit{Evaluation Hallucinations}'', where models optimize linguistic fluency rather than factual clinical correctness, leading to diagnostically critical errors. To bridge this gap, we introduce the \textbf{Clinical Abnormality Benchmarking Substrate (CABS)}, a structured system that decomposes radiology reports into verifiable clinical semantic units. Using CABS, we identify a ``\textit{Mechanistic Divergence}'' in standard RL, where surface-similarity rewards drive policy gradients to bypass medical facts. We therefore propose \textbf{Trajectory-Integral Feedback GRPO (TIF-GRPO)}, a novel framework integrating control-theoretic principles into policy optimization. By formulating clinical reasoning as a pseudo-temporal trajectory for anomaly discovery, TIF-GRPO regulates anatomy-aware rewards via an integral feedback loop that penalizes persistent omissions as cumulative state errors and suppresses hallucinations as excessive control effort. Experiments on 3D CT benchmarks demonstrate that our approach significantly enhances abnormality detection and clinical faithfulness, establishing a new paradigm for fine-grained regulation in medical VLMs. Our project is available at \href{https://github.com/ZJU4HealthCare/TIF-GRPO}{GitHub}.

Konstantin Golobokov, Junyi Chai, Victor Ye Dong et al. · amazon-science

We present DeepGen, a system deployed at web scale for automatically creating sponsored search advertisements (ads) for BingAds customers. We leverage state-of-the-art natural language generation (NLG) models to generate fluent ads from advertiser's web pages in an abstractive fashion and solve practical issues such as factuality and inference speed. In addition, our system creates a customized ad in real-time in response to the user's search query, therefore highlighting different aspects of the same product based on what the user is looking for. To achieve this, our system generates a diverse choice of smaller pieces of the ad ahead of time and, at query time, selects the most relevant ones to be stitched into a complete ad. We improve generation diversity by training a controllable NLG model to generate multiple ads for the same web page highlighting different selling points. Our system design further improves diversity horizontally by first running an ensemble of generation models trained with different objectives and then using a diversity sampling algorithm to pick a diverse subset of generation results for online selection. Experimental results show the effectiveness of our proposed system design. Our system is currently deployed in production, serving ${\sim}4\%$ of global ads served in Bing.

Haogeng Liu, Tao Wang, Jie Cao et al.

Denoising Diffusion Probabilistic Models have shown extraordinary ability on various generative tasks. However, their slow inference speed renders them impractical in speech synthesis. This paper proposes a linear diffusion model (LinDiff) based on an ordinary differential equation to simultaneously reach fast inference and high sample quality. Firstly, we employ linear interpolation between the target and noise to design a diffusion sequence for training, while previously the diffusion path that links the noise and target is a curved segment. When decreasing the number of sampling steps (i.e., the number of line segments used to fit the path), the ease of fitting straight lines compared to curves allows us to generate higher quality samples from a random noise with fewer iterations. Secondly, to reduce computational complexity and achieve effective global modeling of noisy speech, LinDiff employs a patch-based processing approach that partitions the input signal into small patches. The patch-wise token leverages Transformer architecture for effective modeling of global information. Adversarial training is used to further improve the sample quality with decreased sampling steps. We test proposed method with speech synthesis conditioned on acoustic feature (Mel-spectrograms). Experimental results verify that our model can synthesize high-quality speech even with only one diffusion step. Both subjective and objective evaluations demonstrate that our model can synthesize speech of a quality comparable to that of autoregressive models with faster synthesis speed (3 diffusion steps).

Yufei Liu, Jie Cao, Dechang Pi

Social influence prediction has permeated many domains, including marketing, behavior prediction, recommendation systems, and more. However, traditional methods of predicting social influence not only require domain expertise,they also rely on extracting user features, which can be very tedious. Additionally, graph convolutional networks (GCNs), which deals with graph data in non-Euclidean space, are not directly applicable to Euclidean space. To overcome these problems, we extended DeepInf such that it can predict the social influence of COVID-19 via the transition probability of the page rank domain. Furthermore, our implementation gives rise to a deep learning-based personalized propagation algorithm, called DeepPP. The resulting algorithm combines the personalized propagation of a neural prediction model with the approximate personalized propagation of a neural prediction model from page rank analysis. Four social networks from different domains as well as two COVID-19 datasets were used to demonstrate the efficiency and effectiveness of the proposed algorithm. Compared to other baseline methods, DeepPP provides more accurate social influence predictions. Further, experiments demonstrate that DeepPP can be applied to real-world prediction data for COVID-19.

Zhida Zhang, Jie Ma, Zhan Peng et al.

The narrative quality of a video fundamentally determines its perceptual value. Although existing video generation methods can produce visually appealing content, they predominantly rely on sparse conditioning signals such as text prompts or first/last frames, which limits precise control over narrative structure and temporal pacing. In this paper, we propose SmartDirector, a framework that enhances the narrative capacity of video generation models through multiple keyframes. SmartDirector supports flexible generation scenarios including single-shot generation, multi-shot narrative synthesis, and video extension. The framework operates in two stages: Director-Gen generates a low-resolution video conditioned on the provided keyframes, and Director-SR refines the output by exploiting high-resolution keyframes as semantic anchors to recover fine-grained details. To enable robust multi-keyframe training, we construct a data pipeline that curates single-shot and multi-shot sequences from movies. Extensive experiments demonstrate that SmartDirector substantially outperforms existing state-of-the-art approaches. We will release the code to facilitate further research.

Zhimin Li, Shusen Liu, Xin Yu et al.

Deep learning approaches have provided state-of-the-art performance in many applications by relying on large and overparameterized neural networks. However, such networks have been shown to be very brittle and are difficult to deploy on resource-limited platforms. Model pruning, i.e., reducing the size of the network, is a widely adopted strategy that can lead to a more robust and compact model. Many heuristics exist for model pruning, but empirical studies show that some heuristics improve performance whereas others can make models more brittle or have other side effects. This work aims to shed light on how different pruning methods alter the network's internal feature representation and the corresponding impact on model performance. To facilitate a comprehensive comparison and characterization of the high-dimensional model feature space, we introduce a visual geometric analysis of feature representations. We decomposed and evaluated a set of critical geometric concepts from the common adopted classification loss, and used them to design a visualization system to compare and highlight the impact of pruning on model performance and feature representation. The proposed tool provides an environment for in-depth comparison of pruning methods and a comprehensive understanding of how model response to common data corruption. By leveraging the proposed visualization, machine learning researchers can reveal the similarities between pruning methods and redundant in robustness evaluation benchmarks, obtain geometric insights about the differences between pruned models that achieve superior robustness performance, and identify samples that are robust or fragile to model pruning and common data corruption to model pruning and data corruption but also obtain insights and explanations on how some pruned models achieve superior robustness performance.

Puning Yang, Jian Liang, Jie Cao et al.

Out-of-distribution (OOD) detection aims to detect test samples that do not fall into any training in-distribution (ID) classes. Prior efforts focus on regularizing models with ID data only, largely underperforming counterparts that utilize auxiliary outliers. However, data safety and privacy make it infeasible to collect task-specific outliers in advance for different scenarios. Besides, using task-irrelevant outliers leads to inferior OOD detection performance. To address the above issue, we present a new setup called test-time OOD detection, which allows the deployed model to utilize real OOD data from the unlabeled data stream during testing. We propose Adaptive Outlier Optimization (AUTO) which allows for continuous adaptation of the OOD detector. Specifically, AUTO consists of three key components: 1) an in-out-aware filter to selectively annotate test samples with pseudo-ID and pseudo-OOD and ingeniously trigger the updating process while encountering each pseudo-OOD sample; 2) a dynamic-updated memory to overcome the catastrophic forgetting led by frequent parameter updates; 3) a prediction-aligning objective to calibrate the rough OOD objective during testing. Extensive experiments show that AUTO significantly improves OOD detection performance over state-of-the-art methods. Besides, evaluations on complicated scenarios (e.g. multi-OOD, time-series OOD) also conduct the superiority of AUTO.

Nguyen Cong Luong, Shaohan Feng, Nguyen Duc Hai et al.

Reinforcement Learning (RL) has long been a powerful solution to various problems in communication networks. However, traditional RL models still face with several limitations. Not only do they rely on large numbers of interactions with the environment, but they are also limited in terms of modeling long-term relationships and tackling partial observability. In recent years, the Transformer model has demonstrated the ability to enhance RL models, allowing them to overcome these issues. Particularly, the self-attention mechanism within the Transformer enables efficient modeling of long-range dependencies and global correlations, as well as accelerates training processes and handles heterogeneous data modalities. In this paper, we present a comprehensive survey of Transformer-based RL algorithms and their applications in communication networks. Specifically, the paper provides the mathematical background of RL and Transformer architectures, along with insights into key issues such as resource allocation, computation offloading, routing, and trajectory control, and network security. We conclude the paper by discussing challenges, open issues, and notable future research directions, including Transformer-enhanced DRL algorithms for semantic communication and network optimization.

Kewei Wei, Bocheng Hu, Jie Cao et al.

Modern Large Multimodal Models (LMMs) have demonstrated extraordinary ability in static image and single-state spatial-temporal understanding. However, their capacity to comprehend the dynamic changes of objects within a shared spatial context between two distinct video observations, remains largely unexplored. This ability to reason about transformations within a consistent environment is particularly crucial for advancements in the field of spatial intelligence. In this paper, we introduce $M^3-Verse$, a Multi-Modal, Multi-State, Multi-Dimensional benchmark, to formally evaluate this capability. It is built upon paired videos that provide multi-perspective observations of an indoor scene before and after a state change. The benchmark contains a total of 270 scenes and 2,932 questions, which are categorized into over 50 subtasks that probe 4 core capabilities. We evaluate 16 state-of-the-art LMMs and observe their limitations in tracking state transitions. To address these challenges, we further propose a simple yet effective baseline that achieves significant performance improvements in multi-state perception. $M^3-Verse$ thus provides a challenging new testbed to catalyze the development of next-generation models with a more holistic understanding of our dynamic visual world. You can get the construction pipeline from https://github.com/Wal-K-aWay/M3-Verse_pipeline and full benchmark data from https://www.modelscope.cn/datasets/WalKaWay/M3-Verse.

Zhenxuan Fan, Jie Cao, Yang Dai et al.

Chain-of-thought (CoT) prompting improves LLM reasoning but incurs high latency and memory cost due to verbose traces, motivating CoT compression with preserved correctness. Existing methods either shorten CoTs at the semantic level, which is often conservative, or prune tokens aggressively, which can miss task-critical cues and degrade accuracy. Moreover, combining the two is non-trivial due to sequential dependency, task-agnostic pruning, and distribution mismatch. We propose \textbf{CtrlCoT}, a dual-granularity CoT compression framework that harmonizes semantic abstraction and token-level pruning through three components: Hierarchical Reasoning Abstraction produces CoTs at multiple semantic granularities; Logic-Preserving Distillation trains a logic-aware pruner to retain indispensable reasoning cues (e.g., numbers and operators) across pruning ratios; and Distribution-Alignment Generation aligns compressed traces with fluent inference-time reasoning styles to avoid fragmentation. On MATH-500 with Qwen2.5-7B-Instruct, CtrlCoT uses 30.7\% fewer tokens while achieving 7.6 percentage points higher than the strongest baseline, demonstrating more efficient and reliable reasoning. Our code will be publicly available at https://github.com/fanzhenxuan/Ctrl-CoT.

Zilong Li, Jie Cao

Ancient people translated classical Chinese into Japanese by annotating around each character. We abstract this process as sequence tagging tasks and fit them into modern language technologies. The research of this annotation and translation system is a facing low-resource problem. We release this problem by introducing a LLM-based annotation pipeline and construct a new dataset from digitalized open-source translation data. We show that under the low-resource setting, introducing auxiliary Chinese NLP tasks has a promoting effect on the training of sequence tagging tasks. We also evaluate the performance of large language models. They achieve high scores in direct machine translation, but they are confused when being asked to annotate characters. Our method could work as a supplement of LLMs.

Jie Cao, Mandi Luo, Junchi Yu et al.

Generative Adversarial Networks (GANs) typically suffer from overfitting when limited training data is available. To facilitate GAN training, current methods propose to use data-specific augmentation techniques. Despite the effectiveness, it is difficult for these methods to scale to practical applications. In this work, we present ScoreMix, a novel and scalable data augmentation approach for various image synthesis tasks. We first produce augmented samples using the convex combinations of the real samples. Then, we optimize the augmented samples by minimizing the norms of the data scores, i.e., the gradients of the log-density functions. This procedure enforces the augmented samples close to the data manifold. To estimate the scores, we train a deep estimation network with multi-scale score matching. For different image synthesis tasks, we train the score estimation network using different data. We do not require the tuning of the hyperparameters or modifications to the network architecture. The ScoreMix method effectively increases the diversity of data and reduces the overfitting problem. Moreover, it can be easily incorporated into existing GAN models with minor modifications. Experimental results on numerous tasks demonstrate that GAN models equipped with the ScoreMix method achieve significant improvements.

Siyu Xing, Jie Cao, Huaibo Huang et al.

Flow matching as a paradigm of generative model achieves notable success across various domains. However, existing methods use either multi-round training or knowledge within minibatches, posing challenges in finding a favorable coupling strategy for straightening trajectories to few-step generation. To address this issue, we propose a novel approach, Straighter trajectories of Flow Matching (StraightFM). It straightens trajectories with the coupling strategy from the entire distribution level. More specifically, during training, StraightFM creates couplings of images and noise via one diffusion model as a coupling prior to straighten trajectories for few-step generation. Our coupling strategy can also integrate with the existing coupling direction from real data to noise, improving image quality in few-step generation. Experimental results on pixel space and latent space show that StraightFM yields attractive samples within 5 steps. Moreover, our unconditional StraightFM is seamlessly compatible with training-free multimodal conditional generation, maintaining high-quality image generation in few steps.

Chun Bao, Jie Cao, Yaqian Ning et al.

Extensive research works demonstrate that the attention mechanism in convolutional neural networks (CNNs) effectively improves accuracy. Nevertheless, few works design attention mechanisms using large receptive fields. In this work, we propose a novel attention method named Rega-net to increase CNN accuracy by enlarging the receptive field. Inspired by the mechanism of the human retina, we design convolutional kernels to resemble the non-uniformly distributed structure of the human retina. Then, we sample variable-resolution values in the Gabor function distribution and fill these values in retina-like kernels. This distribution allows essential features to be more visible in the center position of the receptive field. We further design an attention module including these retina-like kernels. Experiments demonstrate that our Rega-Net achieves 79.96% top-1 accuracy on ImageNet-1K classification and 43.1% mAP on COCO2017 object detection. The mAP of the Rega-Net increased by up to 3.5% compared to baseline networks.

Jie Cao, Yin Zhang

Extreme multi-label text classification (XMTC) is the task of finding the most relevant subset labels from an extremely large-scale label collection. Recently, some deep learning models have achieved state-of-the-art results in XMTC tasks. These models commonly predict scores for all labels by a fully connected layer as the last layer of the model. However, such models can't predict a relatively complete and variable-length label subset for each document, because they select positive labels relevant to the document by a fixed threshold or take top k labels in descending order of scores. A less popular type of deep learning models called sequence-to-sequence (Seq2Seq) focus on predicting variable-length positive labels in sequence style. However, the labels in XMTC tasks are essentially an unordered set rather than an ordered sequence, the default order of labels restrains Seq2Seq models in training. To address this limitation in Seq2Seq, we propose an autoregressive sequence-to-set model for XMTC tasks named OTSeq2Set. Our model generates predictions in student-forcing scheme and is trained by a loss function based on bipartite matching which enables permutation-invariance. Meanwhile, we use the optimal transport distance as a measurement to force the model to focus on the closest labels in semantic label space. Experiments show that OTSeq2Set outperforms other competitive baselines on 4 benchmark datasets. Especially, on the Wikipedia dataset with 31k labels, it outperforms the state-of-the-art Seq2Seq method by 16.34% in micro-F1 score. The code is available at https://github.com/caojie54/OTSeq2Set.

Yuhe Ding, Jian Liang, Jie Cao et al.

Existing face stylization methods always acquire the presence of the target (style) domain during the translation process, which violates privacy regulations and limits their applicability in real-world systems. To address this issue, we propose a new method called MODel-drIven Face stYlization (MODIFY), which relies on the generative model to bypass the dependence of the target images. Briefly, MODIFY first trains a generative model in the target domain and then translates a source input to the target domain via the provided style model. To preserve the multimodal style information, MODIFY further introduces an additional remapping network, mapping a known continuous distribution into the encoder's embedding space. During translation in the source domain, MODIFY fine-tunes the encoder module within the target style-persevering model to capture the content of the source input as precisely as possible. Our method is extremely simple and satisfies versatile training modes for face stylization. Experimental results on several different datasets validate the effectiveness of MODIFY for unsupervised face stylization.

Zihan Lin, Xiaohan Wang, Jie Cao et al.

Reinforcement Learning with Verifiable Rewards (RLVR) enhances reasoning of Large Language Models (LLMs) but usually exhibits limited generation diversity due to the over-incentivization of positive rewards. Although methods like Negative Sample Reinforcement (NSR) mitigate this issue by upweighting penalty from negative samples, they may suppress the semantic distributions shared between positive and negative responses. To boost reasoning ability without losing diversity, this paper proposes negative sample projection Residual Reinforcement Learning (ResRL) that decouples similar semantic distributions among positive and negative responses. We theoretically link Lazy Likelihood Displacement (LLD) to negative-positive head-gradient interference and derive a single-forward proxy that upper-bounds representation alignment to guide conservative advantage reweighting. ResRL then projects negative-token hidden representations onto an SVD-based low-rank positive subspace and uses projection residuals to modulate negative gradients, improving reasoning while preserving diversity and outperforming strong baselines on average across twelve benchmarks spanning Mathematics, Code, Agent Tasks, and Function Calling. Notably, ResRL surpasses NSR on mathematical reasoning by 9.4\% in Avg@16 and 7.0\% in Pass@128. Code is available at https://github.com/1229095296/ResRL.git.

Chao Jin, Wenkui Yang, Hao Sun et al.

While progress in GUI agents has been largely driven by industrial-scale training, ungrounded hallucinations often trigger cascading failures in real-world deployments.Unlike general VLM domains, the GUI agent field lacks a hallucination-focused suite for fine-grained diagnosis, reliable evaluation, and targeted mitigation.To bridge this gap, we introduce HalluClear, a comprehensive suite for hallucination mitigation in GUI agents as a complement to computation-intensive scaling. HalluClear comprises: (1) a GUI-specific hallucination taxonomy derived from empirical failure analysis; (2) a calibrated three-stage evaluation workflow which enhances VLM-as-a-judge reliability via expert-annotated benchmarking and ensemble credibility estimation; and (3) a mitigation scheme based on closed-loop structured reasoning, enabling lightweight continual post-training with cold-start initialization for both generalist and GUI-specialist agents. Experiments across representative agents and public benchmarks demonstrate that post-training on only 9K samples within our suite can significantly reduce hallucinations, thereby improving grounding and action fidelity, offering a compute-efficient pathway to robust GUI automation.

Jia Li, Jie Cao, JunXian Duan et al.

We propose a new challenging task namely IDentity Stylization (IDS) across heterogeneous domains. IDS focuses on stylizing the content identity, rather than completely swapping it using the reference identity. We use an effective heterogeneous-network-based framework $Styleverse$ that uses a single domain-aware generator to exploit the Metaverse of diverse heterogeneous faces, based on the proposed dataset FS13 with limited data. FS13 means 13 kinds of Face Styles considering diverse lighting conditions, art representations and life dimensions. Previous similar tasks, \eg, image style transfer can handle textural style transfer based on a reference image. This task usually ignores the high structure-aware facial area and high-fidelity preservation of the content. However, Styleverse intends to controllably create topology-aware faces in the Parallel Style Universe, where the source facial identity is adaptively styled via AdaIN guided by the domain-aware and reference-aware style embeddings from heterogeneous pretrained models. We first establish the IDS quantitative benchmark as well as the qualitative Styleverse matrix. Extensive experiments demonstrate that Styleverse achieves higher-fidelity identity stylization compared with other state-of-the-art methods.

Wenkui Yang, Chao Jin, Haisu Zhu et al.

Existing red-teaming studies on GUI agents have important limitations. Adversarial perturbations typically require white-box access, which is unavailable for commercial systems, while prompt injection is increasingly mitigated by stronger safety alignment. To study robustness under a more practical threat model, we propose Semantic-level UI Element Injection, a red-teaming setting that overlays safety-aligned and harmless UI elements onto screenshots to misdirect the agent's visual grounding. Our method uses a modular Editor-Overlapper-Victim pipeline and an iterative search procedure that samples multiple candidate edits, keeps the best cumulative overlay, and adapts future prompt strategies based on previous failures. Across five victim models, our optimized attacks improve attack success rate by up to 4.4x over random injection on the strongest victims. Moreover, elements optimized on one source model transfer effectively to other target models, indicating model-agnostic vulnerabilities. After the first successful attack, the victim still clicks the attacker-controlled element in more than 15% of later independent trials, versus below 1% for random injection, showing that the injected element acts as a persistent attractor rather than simple visual clutter.

Yuzhe Chen, Jie Cao, Youquan Wang et al.

Existing multi-behavior recommendations tend to prioritize performance at the expense of explainability, while current explainable methods suffer from limited generalizability due to their reliance on external information. Neuro-Symbolic integration offers a promising avenue for explainability by combining neural networks with symbolic logic rule reasoning. Concurrently, we posit that user behavior chains inherently embody an endogenous logic suitable for explicit reasoning. However, these observational multiple behaviors are plagued by confounders, causing models to learn spurious correlations. By incorporating causal inference into this Neuro-Symbolic framework, we propose a novel Causal Neuro-Symbolic Reasoning model for Explainable Multi-Behavior Recommendation (CNRE). CNRE operationalizes the endogenous logic by simulating a human-like decision-making process. Specifically, CNRE first employs hierarchical preference propagation to capture heterogeneous cross-behavior dependencies. Subsequently, it models the endogenous logic rule implicit in the user's behavior chain based on preference strength, and adaptively dispatches to the corresponding neural-logic reasoning path (e.g., conjunction, disjunction). This process generates an explainable causal mediator that approximates an ideal state isolated from confounding effects. Extensive experiments on three large-scale datasets demonstrate CNRE's significant superiority over state-of-the-art baselines, offering multi-level explainability from model design and decision process to recommendation results.

Zhi Li, Songkun Yan, Jie Cao et al.

Calibrating distributed hydrologic models is a critical bottleneck across operational water resources management - streamflow prediction, reservoir operation, drought monitoring, infrastructure design, and flood forecasting all depend on it. Each basin demands an expert to translate hydrograph signatures into adjustments of a high-dimensional parameter vector, and the resulting workflow does not transfer between watersheds. We ask: can frontier large language model (LLM) agents replace the human hydrologic modeler, and if not, what would it take? We benchmark nine frontier LLM agents - Claude Opus 4.6/4.7, Sonnet 4.6, GPT-5/5.4/5.4-pro, and Gemini 2.5-pro/3.1-pro/3-flash - on the operational CREST distributed hydrologic model used by the U.S. National Weather Service for flash-flood forecasting. Best-of-twenty-rounds Nash-Sutcliffe Efficiency (NSE) across four held-out gauges spanning 329-40,792 km2 ranges from -0.16 (GPT-5.4) to 0.75 (Sonnet 4.6); the ceiling reproduces across all three vendors and capability tiers, with the strongest models concentrating in the 0.65-0.75 band, and no model reaches the human-expert reference except Opus-4.7 on one gauge. We argue this gap is not a parameter-count problem but a domain-grounding problem. We then propose HYDROAGENT, fine-tuning open-weight Qwen3-4B with supervised fine-tuning on 2,576 expert calibration trajectories and Group-Relative Policy Optimization using NSE as a verifiable reward from online CREST simulations - reinforcement learning with simulation feedback (RLSF). For Earth system science, a small domain-tuned policy with simulator-in-the-loop RL is a more compute-efficient and physically faithful path than scaling generic frontier models, and the multi-modal richness of Earth data - remote sensing, in-situ time series, and forecaster narrative - makes domain agents a leveraged direction for AI in physical science.

Zhanxin Hao, Jie Cao, Ruimiao Li et al.

Multi-agent AI systems, which simulate diverse instructional roles such as teachers and peers, offer new possibilities for personalized and interactive learning. Yet, student-AI interaction patterns and their pedagogical implications remain unclear. This study explores how university students engaged with multiple AI agents, and how these interactions influenced cognitive outcomes (learning gains) and non-cognitive factors (motivation, technology acceptance). Based on MAIC, an online learning platform with multi-agent, the research involved 305 university students and 19,365 lines of dialogue data. Pre- and post-test scores, self-reported motivation and technology acceptance were also collected. The study identified two engagement patterns: co-construction of knowledge and co-regulation. Lag sequential analysis revealed that students with lower prior knowledge relied more on co-construction of knowledge sequences, showing higher learning gains and post-course motivation. In contrast, students with higher prior knowledge engaged more in co-regulation behaviors but exhibited limited learning improvement. Technology acceptance increased across all groups. These findings suggest that multi-agent AI systems can adapt to students' varying needs, support differentiated engagement, and reduce performance gaps. Implications for personalized system design and future research directions are discussed.

Chloe Qianhui Zhao, Jie Cao, Jionghao Lin et al.

Providing timely, targeted, and multimodal feedback helps students quickly correct errors, build deep understanding and stay motivated, yet making it at scale remains a challenge. This study introduces a real-time AI-facilitated multimodal feedback system that integrates structured textual explanations with dynamic multimedia resources, including the retrieved most relevant slide page references and streaming AI audio narration. In an online crowdsourcing experiment, we compared this system against fixed business-as-usual feedback by educators across three dimensions: (1) learning effectiveness, (2) learner engagement, (3) perceived feedback quality and value. Results showed that AI multimodal feedback achieved learning gains equivalent to original educator feedback while significantly outperforming it on perceived clarity, specificity, conciseness, motivation, satisfaction, and reducing cognitive load, with comparable correctness, trust, and acceptance. Process logs revealed distinct engagement patterns: for multiple-choice questions, educator feedback encouraged more submissions; for open-ended questions, AI-facilitated targeted suggestions lowered revision barriers and promoted iterative improvement. These findings highlight the potential of AI multimodal feedback to provide scalable, real-time, and context-aware support that both reduces instructor workload and enhances student experience.

Zelin Zhang, Qi Li, Jie Cao et al.

Generative AI systems are increasingly used not only to produce content but also to retrieve data, invoke tools, and execute actions. This work examines the security and safety implications of that shift across content-level, model-level, and agentic threats. We analyze how attacker access requirements, system autonomy, and the scope of potential harm change as models move from generating artifacts to executing operations through tool chains and external APIs. We then assess technical countermeasures including detection, watermarking, alignment, and emerging agentic safeguards, and show that several depend on forms of institutional coordination that current governance arrangements do not yet provide. Across the cases examined, capability deployment and attack-surface expansion repeatedly outpace defensive responses as systems move from generating content to executing real-world actions.

Zihan Lin, Xiaohan Wang, Hexiong Yang et al.

While reinforcement learning (RL) shows promise in training tool-use large language models (LLMs) using verifiable outcome rewards, existing methods largely overlook the potential of explicit reasoning rewards to bolster reasoning and tool utilization. Furthermore, natively combining reasoning and outcome rewards may yield suboptimal performance or conflict with the primary optimization objective. To address this, we propose advantage-weighted policy optimization (AWPO) -- a principled RL framework that effectively integrates explicit reasoning rewards to enhance tool-use capability. AWPO incorporates variance-aware gating and difficulty-aware weighting to adaptively modulate advantages from reasoning signals based on group-relative statistics, alongside a tailored clipping mechanism for stable optimization. Extensive experiments demonstrate that AWPO achieves state-of-the-art performance across standard tool-use benchmarks, significantly outperforming strong baselines and leading closed-source models in challenging multi-turn scenarios. Notably, with exceptional parameter efficiency, our 4B model surpasses Grok-4 by 16.0 percent in multi-turn accuracy while preserving generalization capability on the out-of-distribution MMLU-Pro benchmark.

Yunzheng Su, Lei Jiang, Jie Cao

In recent years, with the development of computing resources and LiDAR, point cloud semantic segmentation has attracted many researchers. For the sparsity of point clouds, although there is already a way to deal with sparse convolution, multi-scale features are not considered. In this letter, we propose a feature extraction module based on multi-scale sparse convolution and a feature selection module based on channel attention and build a point cloud segmentation network framework based on this. By introducing multi-scale sparse convolution, the network could capture richer feature information based on convolution kernels with different sizes, improving the segmentation result of point cloud segmentation. Experimental results on Stanford large-scale 3-D Indoor Spaces(S3DIS) dataset and outdoor dataset(SemanticKITTI), demonstrate effectiveness and superiority of the proposed mothod.

Cuong Huynh, Jie Cao

This paper describes the OUNLP system submitted to the TSAR-2025 Shared Task (Alva-Manchego et al., 2025), designed for readability-controlled text simplification using LLM-prompting-based generation. Based on the analysis of prompt-based text simplification methods, we discovered an interesting finding that text simplification performance is highly related to the gap between the source CEFR (Arase et al., 2022) level and the target CEFR level. Inspired by this finding, we propose two multi-round simplification methods and generate them via GPT-4o: rule-based simplification (MRS-Rule) and jointly rule-based LLM simplification (MRS-Joint). Our submitted systems ranked 7 out of 20 teams. Later improvements with MRS-Joint show that taking the LLM simplified candidates as the starting point could further boost the multi-round simplification performance.

Wenkui Yang, Zhida Zhang, Xiaoqiang Zhou et al.

The emergence and popularity of facial deepfake methods spur the vigorous development of deepfake datasets and facial forgery detection, which to some extent alleviates the security concerns about facial-related artificial intelligence technologies. However, when it comes to human body forgery, there has been a persistent lack of datasets and detection methods, due to the later inception and complexity of human body generation methods. To mitigate this issue, we introduce TikTok-DeepFake (TT-DF), a novel large-scale diffusion-based dataset containing 6,120 forged videos with 1,378,857 synthetic frames, specifically tailored for body forgery detection. TT-DF offers a wide variety of forgery methods, involving multiple advanced human image animation models utilized for manipulation, two generative configurations based on the disentanglement of identity and pose information, as well as different compressed versions. The aim is to simulate any potential unseen forged data in the wild as comprehensively as possible, and we also furnish a benchmark on TT-DF. Additionally, we propose an adapted body forgery detection model, Temporal Optical Flow Network (TOF-Net), which exploits the spatiotemporal inconsistencies and optical flow distribution differences between natural data and forged data. Our experiments demonstrate that TOF-Net achieves favorable performance on TT-DF, outperforming current state-of-the-art extendable facial forgery detection models. For our TT-DF dataset, please refer to https://github.com/HashTAG00002/TT-DF.

Jie Cao, Chloe Qianhui Zhao, Christian Schunn et al.

Feedback is vital for learning, yet its effectiveness depends not only on its content but also on how it engages students in the learning process. Large Language Models (LLMs) offer novel opportunities to efficiently generate rich, formative feedback, ranging from direct explanations to incrementally layered scaffolding designed to foster learner autonomy. Despite these affordances, it remains unclear whether layered feedback (which sequences encouragement and prompts prior to revealing the correct answer) actually improves engagement and learning outcomes. To address this, we randomly assigned 199 participants to receive either layered or non-layered LLM-generated feedback. We assessed its impact on learning performance, behavioral and cognitive engagement, and affective perceptions, to determine how these factors mediate learning performance. Results indicate that layered feedback elicited slightly higher behavioral engagement and, as anticipated, was perceived as more encouraging and supportive of independence. However, it concurrently induced greater mental effort. Mediation analyses revealed a positive affective pathway driven by perceived encouragement, which was counteracted by a negative behavioral pathway linked to the average number of tasks requiring $\geq 3$ submissions; the cognitive pathway (mental effort) was non-significant. Taken together, layered feedback resulted in significantly poorer learning outcomes compared to non-layered feedback. These findings illuminate a critical trade-off: while layered scaffolding enhances engagement and positive perceptions, it can detrimentally impact actual learning performance. This study contributes nuanced insights for the design of automated, LLM-driven feedback systems by integrating outcome, perception, and mechanism-level analyses.

Junchi Yu, Jie Cao, Ran He

Subgraph recognition aims at discovering a compressed substructure of a graph that is most informative to the graph property. It can be formulated by optimizing Graph Information Bottleneck (GIB) with a mutual information estimator. However, GIB suffers from training instability and degenerated results due to its intrinsic optimization process. To tackle these issues, we reformulate the subgraph recognition problem into two steps: graph perturbation and subgraph selection, leading to a novel Variational Graph Information Bottleneck (VGIB) framework. VGIB first employs the noise injection to modulate the information flow from the input graph to the perturbed graph. Then, the perturbed graph is encouraged to be informative to the graph property. VGIB further obtains the desired subgraph by filtering out the noise in the perturbed graph. With the customized noise prior for each input, the VGIB objective is endowed with a tractable variational upper bound, leading to a superior empirical performance as well as theoretical properties. Extensive experiments on graph interpretation, explainability of Graph Neural Networks, and graph classification show that VGIB finds better subgraphs than existing methods. Code is avaliable at https://github.com/Samyu0304/VGIB

Xiaoxu Li, Jijie Wu, Zhuo Sun et al.

Few-shot learning for fine-grained image classification has gained recent attention in computer vision. Among the approaches for few-shot learning, due to the simplicity and effectiveness, metric-based methods are favorably state-of-the-art on many tasks. Most of the metric-based methods assume a single similarity measure and thus obtain a single feature space. However, if samples can simultaneously be well classified via two distinct similarity measures, the samples within a class can distribute more compactly in a smaller feature space, producing more discriminative feature maps. Motivated by this, we propose a so-called \textit{Bi-Similarity Network} (\textit{BSNet}) that consists of a single embedding module and a bi-similarity module of two similarity measures. After the support images and the query images pass through the convolution-based embedding module, the bi-similarity module learns feature maps according to two similarity measures of diverse characteristics. In this way, the model is enabled to learn more discriminative and less similarity-biased features from few shots of fine-grained images, such that the model generalization ability can be significantly improved. Through extensive experiments by slightly modifying established metric/similarity based networks, we show that the proposed approach produces a substantial improvement on several fine-grained image benchmark datasets. Codes are available at: https://github.com/spraise/BSNet

Xiaoxu Li, Liyun Yu, Xiaochen Yang et al.

Despite achieving state-of-the-art performance, deep learning methods generally require a large amount of labeled data during training and may suffer from overfitting when the sample size is small. To ensure good generalizability of deep networks under small sample sizes, learning discriminative features is crucial. To this end, several loss functions have been proposed to encourage large intra-class compactness and inter-class separability. In this paper, we propose to enhance the discriminative power of features from a new perspective by introducing a novel neural network termed Relation-and-Margin learning Network (ReMarNet). Our method assembles two networks of different backbones so as to learn the features that can perform excellently in both of the aforementioned two classification mechanisms. Specifically, a relation network is used to learn the features that can support classification based on the similarity between a sample and a class prototype; at the meantime, a fully connected network with the cross entropy loss is used for classification via the decision boundary. Experiments on four image datasets demonstrate that our approach is effective in learning discriminative features from a small set of labeled samples and achieves competitive performance against state-of-the-art methods. Codes are available at https://github.com/liyunyu08/ReMarNet.

Xiaoxu Li, Dongliang Chang, Zhanyu Ma et al.

A deep neural network of multiple nonlinear layers forms a large function space, which can easily lead to overfitting when it encounters small-sample data. To mitigate overfitting in small-sample classification, learning more discriminative features from small-sample data is becoming a new trend. To this end, this paper aims to find a subspace of neural networks that can facilitate a large decision margin. Specifically, we propose the Orthogonal Softmax Layer (OSL), which makes the weight vectors in the classification layer remain orthogonal during both the training and test processes. The Rademacher complexity of a network using the OSL is only $\frac{1}{K}$, where $K$ is the number of classes, of that of a network using the fully connected classification layer, leading to a tighter generalization error bound. Experimental results demonstrate that the proposed OSL has better performance than the methods used for comparison on four small-sample benchmark datasets, as well as its applicability to large-sample datasets. Codes are available at: https://github.com/dongliangchang/OSLNet.

Yingheng Tang, Wenbin Xu, Jie Cao et al.

Understanding and predicting the properties of inorganic materials is crucial for accelerating advancements in materials science and driving applications in energy, electronics, and beyond. Integrating material structure data with language-based information through multi-modal large language models (LLMs) offers great potential to support these efforts by enhancing human-AI interaction. However, a key challenge lies in integrating atomic structures at full resolution into LLMs. In this work, we introduce MatterChat, a versatile structure-aware multi-modal LLM that unifies material structural data and textual inputs into a single cohesive model. MatterChat employs a bridging module to effectively align a pretrained machine learning interatomic potential with a pretrained LLM, reducing training costs and enhancing flexibility. Our results demonstrate that MatterChat significantly improves performance in material property prediction and human-AI interaction, surpassing general-purpose LLMs such as GPT-4. We also demonstrate its usefulness in applications such as more advanced scientific reasoning and step-by-step material synthesis.

Jie Cao, Abhijit Suresh, Jennifer Jacobs et al.

Human tutoring interventions play a crucial role in supporting student learning, improving academic performance, and promoting personal growth. This paper focuses on analyzing mathematics tutoring discourse using talk moves - a framework of dialogue acts grounded in Accountable Talk theory. However, scaling the collection, annotation, and analysis of extensive tutoring dialogues to develop machine learning models is a challenging and resource-intensive task. To address this, we present SAGA22, a compact dataset, and explore various modeling strategies, including dialogue context, speaker information, pretraining datasets, and further fine-tuning. By leveraging existing datasets and models designed for classroom teaching, our results demonstrate that supplementary pretraining on classroom data enhances model performance in tutoring settings, particularly when incorporating longer context and speaker information. Additionally, we conduct extensive ablation studies to underscore the challenges in talk move modeling.

Jie Cao, Zhanxin Hao, Jifan Yu

Educational dialogue is critical for decoding student learning processes, yet manual annotation remains time-consuming. This study evaluates the efficacy of GPT-5.2 and Gemini-3 using three prompting strategies (few-shot, single-agent, and multi-agent reflection) across diverse subjects, educational levels, and four coding dimensions. Results indicate that while multi-agent prompting achieved the highest accuracy, the results did not reach statistical significance. Accuracy proved highly context-dependent, with significantly higher performance in K-12 datasets compared to university-level data, alongside disciplinary variations within the same educational level. Performance peaked in the affective dimension but remained lowest in the cognitive dimension. Furthermore, analysis revealed four bias patterns: (1) Gemini-3 exhibited a consistent optimistic bias in the affective dimension across all subjects; (2) the cognitive dimension displayed domain-specific directional bias, characterized by systematic underestimation in Mathematics versus overestimation in Psychology; (3) both models are more prone to overestimation than underestimation within the meta-cognitive dimension; and (4) behavioral categories such as question, negotiation, and statements were frequently misclassified. These results underscore the need for context-sensitive deployment and targeted mitigation of directional biases in automated annotation.

Jie Cao, Ha Nguyen, Selim Yavuz et al.

Large Language Model (LLM) simulations, where LLMs act as students with varying approaches to learning tasks, can support teachers' noticing of student thinking. However, simulations using zero- or few-shot prompting often yield inauthentic knowledge and language, directing teachers to unrealistic reasoning. We evaluate three approaches (Fine-tuning, Multi-agent, and Direct Preference Optimization; DPO) to improve the authenticity and pedagogical utility of simulated students. All approaches improve cognitive and linguistic authenticity, compared with few-shot prompts. Interviews with elementary mathematics pre-service teachers and researchers (\textit{n} = 8) reveal distinct pedagogical affordances. The fine-tuned model produces realistic, brief responses but limits opportunities to extend students' thinking. Meanwhile, the multi-agent and DPO approaches generate explicit reasoning behind student strategies. We discuss implications for designing LLM simulations that balance authenticity with instructional utility for teacher learning.

Zihan Lin, Xiaohan Wang, Jie Cao et al.

Large language models (LLMs) transcend passive generation and act as goal-directed agents by invoking external tools. Reinforcement learning (RL) offers a principled framework for optimizing these emergent tool-use policies, yet the prevailing paradigm relies exclusively on sparse outcome rewards and lacks consideration of the particularity of tool-use tasks, inflating policy-gradient variance and resulting in inefficient training. To better understand and address these challenges, we first establish a theoretical link between policy entropy and training stability of tool-use tasks, which reveals that structured, low-entropy tokens are primary determinants of rewards. Motivated by this insight, we propose \textbf{Res}haped \textbf{T}oken-level policy gradients (\textbf{ResT}) for tool-use tasks. ResT reshapes the policy gradient through entropy-informed token reweighting, progressively upweighting reasoning tokens as training proceeds. This entropy-aware scheme enables a smooth shift from structural correctness to semantic reasoning and stabilizes convergence in multi-turn tool-use tasks. Evaluation on BFCL and API-Bank shows that ResT achieves state-of-the-art results, outperforming prior methods by up to $8.76\%$. When fine-tuned on a 4B base LLM, ResT further surpasses GPT-4o by $4.11\%$ on single-turn tasks and $1.50\%$ on multi-turn base tasks.

Jayanth Krishna Chundru, Rudrashis Poddar, Jie Cao et al.

We investigate whether large language models encode latent knowledge of frame semantics, focusing on frame identification, a core challenge in frame semantic parsing that involves selecting the appropriate semantic frame for a target word in context. Using the FrameNet lexical resource, we evaluate models under prompt-based inference and observe that they can perform frame identification effectively even without explicit supervision. To assess the impact of task-specific training, we fine-tune the model on FrameNet data, which substantially improves in-domain accuracy while generalizing well to out-of-domain benchmarks. Further analysis shows that the models can generate semantically coherent frame definitions, highlighting the model's internalized understanding of frame semantics.

Zhida Zhang, Jie Cao, Wenkui Yang et al.

The transformer networks are extensively utilized in face forgery detection due to their scalability across large datasets.Despite their success, transformers face challenges in balancing the capture of global context, which is crucial for unveiling forgery clues, with computational complexity.To mitigate this issue, we introduce Band-Attention modulated RetNet (BAR-Net), a lightweight network designed to efficiently process extensive visual contexts while avoiding catastrophic forgetting.Our approach empowers the target token to perceive global information by assigning differential attention levels to tokens at varying distances. We implement self-attention along both spatial axes, thereby maintaining spatial priors and easing the computational burden.Moreover, we present the adaptive frequency Band-Attention Modulation mechanism, which treats the entire Discrete Cosine Transform spectrogram as a series of frequency bands with learnable weights.Together, BAR-Net achieves favorable performance on several face forgery datasets, outperforming current state-of-the-art methods.

Siyu Zha, Weijing Liu, Fei Qin et al.

Generative artificial intelligence (GenAI) is increasingly embedded in computer-supported collaborative learning (CSCL), yet little empirical research has unpacked how different configurations of AI participation reshape collaborative processes. This study investigates how GenAI configuration shapes collaborative regulation in authentic classroom settings. Two eighth-grade classes engaged in small-group creative problem-solving under two conditions: a shared-AI configuration, in which each group interacted with a single AI mentor, and an individual-AI configuration, in which each student accessed a personal AI instance. Using multi-layer discourse coding combined with lag sequential analysis (LSA) and ordered network analysis (ONA), we examined interaction distribution, AI-student coupling, shared regulation processes, and teacher orchestration. Results reveal distinct regulatory dynamics across configurations. Shared AI access promoted convergence-oriented collaboration, with stronger alignment of shared regulatory states and more coordinated group-level reasoning. In contrast, individual AI access distributed support across learners, producing more exploratory and evaluative cycles but also more fragmented interaction patterns, accompanied by increased teacher intervention to manage divergence. These findings suggest that AI configuration functions as a structural design variable that reorganizes the regulatory ecology of classroom collaboration.