Large-scale learner-task interaction data are crucial for intelligent educational systems but are costly to collect and constrained by privacy and learner engagement. Learner simulators play a critical role in simulating scalable learner behavior without the need for continuous involvement of real learners. However, existing methods are predominantly \textbf{individual-centric}, pairing a simulator with each learner to iteratively infer latent knowledge states from dense interaction histories, which is both data- and computation-intensive, and fragile in cold-start scenarios. We propose a \textbf{cohort-aware roll-call simulation paradigm} that first constructs cohort-level proficiency priors and refines individual learner states through a small number of targeted diagnostic queries. Based on this paradigm, we introduce \textbf{Edu-Theater}, an LLM-powered agent system that performs cohort-aware learner simulation via a teacher agent and retrospective roll-call probing over learner logs. Edu-Theater enables scalable future behavior simulation without the need for dense per-learner histories. Experiments on two real-world datasets demonstrate that Edu-Theater achieves higher simulation accuracy with significantly fewer LLM calls, producing synthetic data that enhances downstream applications such as adaptive testing.

Aspect-based sentiment analysis (ABSA) is a fine-grained sentiment classification task. Despite significant improvements in this field, progress is hindered by challenges such as the sparsity of contexts for specific aspects, interference from irrelevant words within sentences, and a lack of research on leveraging correlations between samples. In order to address these issues, we propose a novel framework MASS, Mitigating Aspect-oriented Semantic Sparsity for ABSA. Firstly, we employ advanced prompting techniques with Large Language Models (LLMs) to generate nuanced aspect-specific descriptions, thereby enhancing contexts related to the aspect. Subsequently, we design a novel fusion module aimed at seamlessly integrating aspectual insights with the original sentence. Finally, we develop three pioneering contrastive learning strategies to explore and learn complex correlations between samples, which are crucial for fine-grained sentiment analysis. Experiments on six benchmark datasets demonstrate that our MASS substantially outperforms state-of-the-art techniques and provides valuable insights for applying LLMs to downstream tasks.

Legal Rationale Generation aims to produce coherent and logically rigorous explanations for judicial decisions based on case facts. Recently, Large Language Models (LLMs) have been explored to improve the performance of generation. While LLMs show promise, existing methods struggle to achieve satisfactory results. For example, most approaches rely on fine-tuning LLMs, making them difficult to implement in resource-constrained scenarios. Besides, directly prompting LLMs to generate rationales often results in a lack of legal grounding due to insufficient legal expertise. To this end, in this paper, we propose a Precedent Retrieval-based In-Context Learning (PRICL) for legal rationale generation framework. This framework retrieves precedents that are prior cases with similar factual descriptions, and then uses them as demonstrations to guide LLMs in generating legal rationales. However, PRICL may still face the shortcut learning problem, where the model relies on spurious correlations rather than deeply understanding the logical relationships between case facts and legal rationales in demonstrations. To mitigate this, we further introduce Meta-PRICL, a method that leverages LLMs to autonomously generate and refine meta-reasoning paths for rationale generation. Extensive experimental results on a real-world dataset clearly validate the effectiveness of our proposed method. Code is available at https://github.com/yuelinan/Codes-of-PRICL .

Rapid advances in image generation are eroding the evidentiary value of visual content in settings where authenticity can affect public safety and personal reputation. Yet existing detection benchmarks rarely examine synthetic images in public- and individual-safety contexts, where misleading visual content may carry substantial risks. Here we introduce SafeIMG, a safety-oriented benchmark spanning 12 public- and individual-safety scenarios generated using GPT Image 2. Unlike benchmarks centred on generic imagery and image-level labels, SafeIMG evaluates not only whether detectors recognise synthetic images, but also whether their decisions reflect human-identified anomalies. To this end, SafeIMG provides human annotations that localise suspicious regions and explain local artefacts and higher-level commonsense or physical inconsistencies. We evaluate specialized synthetic-image detectors and vision-language models (VLMs), and find that neither provides reliable detection. The strongest VLM identifies only 49.5% of generated images, whereas the best specialised detector identifies 33.1%, compared with 81.7% accuracy for human evaluators. Model explanations cover only 29.8\% of human-annotated anomalies and predominantly capture local defects in text, faces and hands. Their coverage falls to 15.0% for commonsense conflicts and 12.0% for physical inconsistencies, while detection performance deteriorates further after dissemination-induced image degradation. These findings show that current detectors lack the accuracy, explanatory alignment and robustness needed to evaluate AI-generated images reliably across public- and individual-safety settings.

Long chains of thought (Long CoTs) are widely employed in multimodal reasoning models to tackle complex tasks by capturing detailed visual information. However, these Long CoTs are often excessively lengthy and contain redundant reasoning steps, which can hinder inference efficiency. Compressing these long CoTs is a natural solution, yet existing approaches face two major challenges: (1) they may compromise the integrity of visual-textual reasoning by removing essential alignment cues, and (2) the compression process lacks explainability, making it difficult to discern which information is critical. To address these problems, we propose XMCC, an eXplainable Multimodal CoT Compressor that formulates compression as a sequential decision-making process optimized via reinforcement learning. XMCC can effectively shorten reasoning trajectories while preserving key reasoning steps and answer correctness, and simultaneously generates natural-language explanations for its compression decisions. Extensive experiments on representative multimodal reasoning benchmarks demonstrate that XMCC not only reduces reasoning length but also provides explainable explanations, validating its effectiveness.

Addressing distribution shifts between training and testing sets is paramount for Graph Neural Networks (GNNs). Graph rationale learning has emerged as an effective paradigm to tackle this challenge by extracting invariant subgraphs as rationales for prediction. However, existing methods typically rely on the subgraphs produced by a rationale extractor without explicitly assessing their reliability, particularly during early training stages when the extractor is prone to generating inaccurate or biased rationales. Learning with such unreliable rationales can mislead the classifier and further deteriorate rationale quality, ultimately undermining generalization performance. In this paper, we rethink the objective of graph rationale learning and propose Cooperative Rationale Selection (CoRas), a novel architecture designed to enhance the reliability and robustness of rationale representations. Specifically, CoRas employs a dual-classifier framework to cooperatively select reliable rationales. Motivated by the small-loss criterion, CoRas prioritizes rationales with low prediction loss, while simultaneously selecting rationales with high prediction divergence to mitigate selection bias caused by relying solely on self-predictions during training. This allows CoRas to effectively select reliable rationales for downstream tasks, subsequently enhancing graph rationale learning. Experimental results on both real-world benchmarks and synthetic datasets demonstrate the effectiveness of CoRas in extracting reliable rationale subgraphs for combating out-of-distribution graph settings in graph classification. Code is released at https://github.com/fannazya/CoRas.

Ensemble learning is a powerful approach to improving model performance in addressing big data challenges. It first trains multiple base models with existing machine learning algorithms and then combines their outputs to yield the final prediction result. Despite the demonstrated potential of current ensemble strategies, their integration primarily relies on unidimensional metrics for each base model (e.g., prediction accuracy), which are too coarse-grained to adequately represent the multifaceted abilities of models. In this paper, we propose MADE, a novel Multidimensional Ability aware Dynamic Ensemble paradigm by drawing upon the fine-grained and well-developed measurement of human learning. Specifically, MADE incorporates a dynamic ensemble algorithm that modifies the models' weights in accordance with their evolving abilities throughout the training process. To evaluate the multidimensional abilities of the base models, we develop a diagnostic module that captures individual base models' latent knowledge levels. Besides, an ensemble weight inductor is designed in MADE to generate individual ensemble weights for each sample, different from previous ensemble methods that assign the same ensemble weights to all samples. Extensive experiments on diverse datasets demonstrate the effectiveness of our proposed model in achieving improved ensemble performance.

Tool-Integrated Reasoning (TIR) enables large language models (LLMs) to solve complex tasks by interacting with external tools, yet existing approaches depend on high-quality synthesized trajectories selected by scoring functions and sparse outcome-based rewards, providing limited and biased supervision for learning TIR. To address these challenges, in this paper, we propose AutoTraj, a two-stage framework that automatically learns TIR by repairing and rewarding tool-use trajectories. Specifically, in the supervised fine-tuning (SFT) stage, AutoTraj generates multiple candidate tool-use trajectories for each query and evaluates them along multiple dimensions. High-quality trajectories are directly retained, while low-quality ones are repaired using a LLM (i.e., LLM-as-Repairer). The resulting repaired and high-quality trajectories form a synthetic SFT dataset, while each repaired trajectory paired with its original low-quality counterpart constitutes a dataset for trajectory preference modeling. In the reinforcement learning (RL) stage, based on the preference dataset, we train a trajectory-level reward model to assess the quality of reasoning paths and combine it with outcome and format rewards, thereby explicitly guiding the optimization toward reliable TIR behaviors. Experiments on real-world benchmarks demonstrate the effectiveness of AutoTraj in TIR.

The remarkable success of deep neural networks has sparked growing interest in methods to explain their predictions, with rationale extraction emerging as a key approach. Rationale extraction identifies critical text segments from input data that justify and explain the model’s results, making it especially useful in high-stakes fields. However, strict data privacy regulations in these fields often prevent the sharing of sensitive information. Federated Learning (FL) offers a promising solution by facilitating collaborative model training without requiring data exchange. However, the challenge of heterogeneous data across clients remains a significant barrier to optimal model performance. To this end, in this paper, we introduce Federated Rationale Extraction with Information Competition (FedIC), a novel approach designed for FL settings. Specifically, for each client, FedIC leverages an environment projector to map local data to a global distribution, generating counterfactual samples. Then, from the perspective of information competition, two optimization criteria are designed to ensure that the generated samples remain task-relevant and consistent with the global data distribution, addressing the issue of data heterogeneity. Experimental results on real-world datasets demonstrate the effectiveness of FedIC.

Legal Judgment Prediction (LJP) has emerged as a fundamental aspect of legal AI systems, encompassing both civil and criminal cases. In this paper, we focus on civil LJP, which aims to predict the judgment result based on both case facts and the plaintiffs claim. Existing approaches rely on post-hoc facts summarized by judges, which has proposed leveraging actual fact (i.e., court debate data) to make judgment prediction. However, this approach still suffers from limitations (e.g., data noise), resulting in suboptimal predictions. To this end, we propose a post-hoc facts augmented LJP (PF-LJP) to further explore this "real pattern". Specifically, we encourage the court debate data to be closer to its corresponding post-hoc fact in the vector space with the contrastive learning and distribution alignment method. During the inference, only the court debate data is employed to yield results without any post-hoc fact. Experimental results on a real-world dataset validate the effectiveness of PF-LJP.

In digital education, Cognitive Diagnosis (CD) is essential for modeling learners' cognitive states, such as problem-solving ability and knowledge proficiency, by analyzing their response data, like answer correctness. However, traditional CD methods struggle with \textit{effectiveness} and \textit{efficiency}. They fail to capture the diversity and uncertainty of learners' cognitive states. Additionally, response prediction can be time-consuming. To address these issues, we propose BoxCD, a contrastive probabilistic box embedding model for cognitive diagnosis. BoxCD utilizes high-dimensional axis-aligned hyper-rectangles (boxes) to represent learners and exercises, with the volume of intersecting boxes used to predict learners' responses. This approach effectively captures semantic diversity and uncertainty while enhancing diagnostic effectiveness. To stabilize box embeddings, we integrate contrastive learning objectives with response prediction goals, optimizing the distance between positive and negative samples of learner and exercise boxes to improve uniformity. Additionally, we develop a rank-based response prediction method that leverages the geometric properties of box embeddings to efficiently assess learners' response correctness. Comprehensive experiments on two real-world datasets demonstrate that BoxCD outperforms traditional CD models in both effectiveness and efficiency, showcasing its potential to enhance personalized learning in digital education platforms.

Personalized learning represents a promising educational strategy within intelligent educational systems, aiming to enhance learners' practice efficiency. However, the scarcity of offline practice response data (e.g., answer correctness) and potential biases in human online practice create a significant gap between offline metrics and the actual online performance of personalized learning services. To address this challenge, we introduce Agent4Edu, a novel personalized learning simulator leveraging recent advancements in human intelligence through large language models (LLMs). Agent4Edu features LLM-powered generative agents equipped with learner profile, memory, and action modules tailored to personalized learning algorithms. The learner profiles are initialized using real-world response data, capturing practice styles and cognitive factors. Inspired by psychology theory, the memory module records practice facts and high-level summaries, integrating reflection mechanisms. The action module supports various behaviors, including exercise understanding, analysis, and response generation. Each agent can interact with personalized learning algorithms, such as computerized adaptive testing, enabling a multifaceted evaluation and enhancement of customized services. Through a comprehensive assessment, we explore the strengths and weaknesses of Agent4Edu, emphasizing the consistency and discrepancies in responses between agents and human learners.

Graph In-Context Learning, with the ability to adapt pre-trained graph models to novel and diverse downstream graphs without updating any parameters, has gained much attention in the community. The key to graph in-context learning is to perform downstream graphs conditioned on chosen prompt examples. Existing methods randomly select subgraphs or edges as prompts, leading to noisy graph prompts and inferior model performance. Additionally, due to the gap between pre-training and testing graphs, when the number of classes in the testing graphs is much greater than that in the training, the in-context learning ability will also significantly deteriorate. To tackle the aforementioned challenges, we develop a multi-stage adaptive prompt optimization method GraphPrompter, which optimizes the entire process of generating, selecting, and using graph prompts for better in-context learning capabilities. Firstly, Prompt Generator introduces a reconstruction layer to highlight the most informative edges and reduce irrelevant noise for graph prompt construction. Furthermore, in the selection stage, Prompt Selector employs the k-nearest neighbors algorithm and pre-trained selection layers to dynamically choose appropriate samples and minimize the influence of irrelevant prompts. Finally, we leverage a Prompt Augmenter with a cache replacement strategy to enhance the generalization capability of the pre-trained model on new datasets. Extensive experiments show that GraphPrompter effectively enhances the in-context learning ability of graph models. On average across all the settings, our approach surpasses the state-of-the-art baselines by over 8 released at https://github.com/karin0018/GraphPrompter.

The success of Graph Neural Networks (GNNs) in graph classification has heightened interest in explainable GNNs, particularly through graph rationalization. This method aims to enhance GNNs explainability by identifying subgraph structures (i.e., rationales) that support model predictions. However, existing methods often rely on centralized datasets, posing challenges in scenarios where data privacy is crucial, such as in molecular property prediction. Federated Learning (FL) offers a solution by enabling collaborative model training without sharing raw data. In this context, Federated Graph Rationalization emerges as a promising research direction. However, in each client, the rationalization methods often rely on client-specific shortcuts to compose rationales and make task predictions. Data heterogeneity, characterized by non-IID data across clients, exacerbates this problem, leading to poor prediction performance. To address these challenges, we propose the Environment-aware Data Augmentation (EaDA) method for Federated Graph Rationalization. EaDA comprises two main components: the Environment-aware Rationale Extraction (ERE) module and the Local-Global Alignment (LGA) module. The ERE module employs prototype learning to infer and share abstract environment information across clients, which are then aggregated to form a global environment. This information is used to generate counterfactual samples for local clients, enhancing the robustness of task predictions. The LGA module uses contrastive learning methods to align local and global rationale representations, mitigating performance degradation due to data heterogeneity. Comprehensive experiments on benchmark datasets demonstrate the effectiveness of our approaches.

The pursuit of model explainability has prompted the selective rationalization (aka, rationale extraction) which can identify important features (i.e., rationales) from the original input to support prediction results. Existing methods typically involve a cascaded approach with a selector responsible for extracting rationales from the input, followed by a predictor that makes predictions based on the selected rationales. However, these approaches often neglect the information contained in the non-rationales, underutilizing the input. Therefore, in our prior work, we introduce the Disentanglement-Augmented Rationale Extraction (DARE) method, which disentangles the input into rationale and non-rationale components, and enhances rationale representations by minimizing the mutual information between them. While DARE demonstrates strong performance in rationalization, it may still rely on shortcuts in the training distribution, leading to unfaithful rationales. To this end, in this paper, we propose Faith-DARE, an extension of DARE that aims to extract more reliable rationales by mitigating shortcut dependencies. Specifically, we treat the non-rationale features identified by DARE as environments that are decorrelated from the predictions. By shuffling and recombining these environments with rationales, we generate counterfactual samples and identify invariant rationales that remain predictive across shifted distributions. Extensive experiments on graph and textual datasets validate the effectiveness of Faith-DARE.

Recently, Large Reasoning Models (LRMs) have gradually become a research hotspot due to their outstanding performance in handling complex tasks. Among them, DeepSeek R1 has garnered significant attention for its exceptional performance and open-source nature, driving advancements in the research of R1-style LRMs. Unlike traditional Large Language Models (LLMs), these models enhance logical deduction and decision-making capabilities during reasoning by incorporating mechanisms such as long chain-of-thought and self-reflection through reinforcement learning. However, with the widespread application of these models, the problem of overthinking has gradually emerged. Specifically, when generating answers, these models often construct excessively long reasoning chains with redundant or repetitive steps, which leads to reduced reasoning efficiency and may affect the accuracy of the final answer. To this end, various efficient reasoning methods have been proposed, aiming to reduce the length of reasoning paths without compromising model performance and reasoning capability. By reviewing the current research advancements in the field of efficient reasoning methods systematically, we categorize existing works into two main directions based on the lens of single-model optimization versus model collaboration: (1) Efficient Reasoning with Single Model, which focuses on improving the reasoning efficiency of individual models; and (2) Efficient Reasoning with Model Collaboration, which explores optimizing reasoning paths through collaboration among multiple models. Besides, we maintain a public GitHub repository that tracks the latest progress in efficient reasoning methods.

Programming Knowledge Tracking (PKT) aims to dynamically diagnose learners' mastery levels of programming knowledge based on their coding activities, facilitating more effective and personalized programming education. However, current PKT studies primarily focus on the implicit relationship between code content and knowledge assessment, often overlooking two types of noise signals in long-term programming activities: unwanted signals from unrelated submissions and weak signals from minor modifications. This practical challenge significantly limits model performance and application. To address this issue, we propose Coda, a Code graph-based tuning adaptor designed to enhance existing PKT models by identifying and mitigating the impact of noise. Specifically, Coda first transforms the loose code sequences submitted by each learner into a compact code graph. By leveraging this code graph, unwanted signals can be identified from a semantic similarity perspective. We then apply a cluster-aware GCN to the code graph, which improves the discrimination of weak signals and enables their clustering for identification. Finally, a lightweight yet effective adaptor is incorporated into the PKT task through optimization with two noise feature-based constraints and a navigational regularization term, to correct knowledge states affected by noise. It is worth mentioning that the Coda framework is model-agnostic and can be adapted to most existing PKT solutions. Extensive experimental results on four real-world datasets demonstrate that Coda effectively performs the PKT task in the presence of noisy programming records, outperforming typical baselines.

Graph Neural Networks (GNNs) have demonstrated remarkable performance in graph classification tasks. However, ensuring the explainability of their predictions remains a challenge. To address this, graph rationalization methods have been introduced to generate concise subsets of the original graph, known as rationales, which serve to explain the predictions made by GNNs. Existing rationalizations often rely on shortcuts in data for prediction and rationale composition. In response, de-shortcut rationalization methods have been proposed, which commonly leverage counterfactual augmentation to enhance data diversity for mitigating the shortcut problem. Nevertheless, these methods have predominantly focused on centralized datasets and have not been extensively explored in the Federated Learning (FL) scenarios. To this end, in this paper, we propose a Federated Graph Rationalization (FedGR) with anti-shortcut augmentations to achieve self-explaining GNNs, which involves two data augmenters. These augmenters are employed to produce client-specific shortcut conflicted samples at each client, which contributes to mitigating the shortcut problem under the FL scenarios. Experiments on real-world benchmarks and synthetic datasets validate the effectiveness of FedGR under the FL scenarios.

Recent advancements in Graph Neural Networks (GNNs) have spurred an upsurge of research dedicated to enhancing the explainability of GNNs, particularly in critical domains such as medicine. A promising approach is the self-explaining method, which outputs explanations along with predictions. However, existing self-explaining models require a large amount of training data, rendering them unavailable in few-shot scenarios. To address this challenge, in this paper, we propose a Meta-learned Self-Explaining GNN (MSE-GNN), a novel framework that generates explanations to support predictions in few-shot settings. MSE-GNN adopts a two-stage self-explaining structure, consisting of an explainer and a predictor. Specifically, the explainer first imitates the attention mechanism of humans to select the explanation subgraph, whereby attention is naturally paid to regions containing important characteristics. Subsequently, the predictor mimics the decision-making process, which makes predictions based on the generated explanation. Moreover, with a novel meta-training process and a designed mechanism that exploits task information, MSE-GNN can achieve remarkable performance on new few-shot tasks. Extensive experimental results on four datasets demonstrate that MSE-GNN can achieve superior performance on prediction tasks while generating high-quality explanations compared with existing methods. The code is publicly available at https://github.com/jypeng28/MSE-GNN.