Conventional feedback, even when accompanied by brief explanations, rarely uncovers the hidden contradictions that trigger a learner's mistake. We bridge this gap with counterfactual question generation (CFQG): given a learner's answer, generate a follow-up question that deliberately contradicts it, compelling the learner to confront the underlying conflict. CFQG thus transforms assessment from passive scoring into an interactive and contradiction-centered dialogue that supports knowledge repair. To automate CFQG, we propose GapProbe, which probes the knowledge gap between a learner’s belief and curated facts through a knowledge graph (KG), then designs counterfactual questions (CFQs) that negate the belief. Identifying contradiction-aware triples, and more importantly, selecting those most likely to confuse the learner, are highly challenging in large-scale KGs. GapProbe tackles these challenges with an iterative ProConB cycle coupled with a schema-aware KGMap. By caching one- and multi-hop schema patterns of the KG, KGMap provides ``roadmap'' to guide LLMs jump to deep and contradiction-aware triples, beyond traditional step-wise graph traversal. We present the CFQG benchmark and corresponding metrics for evaluating how generated CFQs trigger, focus, and deepen learner reflection through explicit contradictions. Experiments on multiple datasets and LLMs show that GapProbe boosts LLM reasoning over KGs and generates follow-up questions that consistently promote deeper and more focused learner reflection.

Drug repositioning (DR), identifying new uses for approved drugs, accelerates drug discovery. To address label sparsity in inferring drug-disease associations (DDAs), we propose DCDR, a heterogeneous graph contrastive learning method with diffusion and causal representation. DCDR resolves two key issues in computational DR: 1) Semantic degradation in contrastive views: Standard random perturbations damage pharmacological relationships. Our diffusion paradigm generates valid variations via structured noise and fidelity-driven reconstruction, preserving interactions while boosting diversity. 2) Confounding bias in representations: Protein-mediated spurious correlations distort embeddings. Our causal framework eliminates this by separating direct therapeutic effects from confounding paths through protein intervention, counterfactual reasoning, and adaptive fusion, isolating deconfounded semantics.$\mathbf{1 0}$-fold cross-validation on three benchmarks shows DCDR outperforms state-of-the-art methods significantly. A case study confirms its ability to identify biologically plausible candidates for Alzheimer's disease.

The Direct Segment Anything Model (DirectSAM) excels in class-agnostic contour extraction. In this paper, we explore its use by applying it to optical remote sensing imagery, where semantic contour extraction—such as identifying buildings, road networks, and coastlines-holds significant practical value. Those applications are currently handled via training specialized small models separately on small datasets in each domain. We introduce a foundation model derived from DirectSAM, termed DirectSAM-RS, which not only inherits the strong segmentation capability acquired from natural images, but also benefits from a large-scale dataset we created for remote sensing semantic contour extraction. This dataset comprises over 34k image-text-contour triplets, making it at least 30 times larger than individual dataset. DirectSAM-RS integrates a prompter module: a text encoder and cross-attention layers attached to the DirectSAM architecture, which allows flexible conditioning on target class labels or referring expressions. We evaluate the DirectSAM-RS in both zero-shot and fine-tuning setting, and demonstrate that it achieves state-of-the-art performance across several downstream benchmarks.

Semi-supervised semantic segmentation has attracted considerable attention for its ability to mitigate the reliance on extensive labeled data. However, existing consistency regularization methods only utilize high certain pixels with prediction confidence surpassing a fixed threshold for training, failing to fully leverage the potential supervisory information within the network. Therefore, this paper proposes the Uncertainty-participation Context Consistency Learning (UCCL) method to explore richer supervisory signals. Specifically, we first design the semantic backpropagation update (SBU) strategy to fully exploit the knowledge from uncertain pixel regions, enabling the model to learn consistent pixel-level semantic information from those areas. Furthermore, we propose the class-aware knowledge regulation (CKR) module to facilitate the regulation of class-level semantic features across different augmented views, promoting consistent learning of class-level semantic information within the encoder. Experimental results on two public benchmarks demonstrate that our proposed method achieves state-of-the-art performance. Our code is available at https://github.com/YUKEKEJAN/UCCL.

Recent researches have yielded promising results by integrating online action detection and action anticipation tasks to explore the correlations between past, present and future. However, these approaches treat incomplete historical information equally and neglect intrinsic connections between actions, resulting in a limited perception of the throughout evolution. To address this limitation, we reconsider the patterns and dependencies in event evolution, innovatively constructing a comprehensive deductive process that inscribes the entire temporal spectrum via procedural features. Here, we propose the Throughout Procedural Transformer (TPT) comprising Procedural History Evolution Encoder and Progressive Deduction Decoder, to thoroughly span the entirety of time from history to the future through procedural modeling. TPT utilizes long-term procedural history acquired through procedure sampling to model long-term procedural future, thereby enhancing cognitive inference ability by enriching short-term history and short-term future with a broad grasp of throughout event evolution. We conduct extensive experiments to evaluate TPT on five demanding benchmarks THUMOS’14, TVSeries, FineAction, HACS and EPIC-Kitchens-100 for online action detection and anticipation tasks, demonstrating significant improvements over existing methods.

With the rapid advancement of deep learning technologies, self-supervised learning utilizing large-scale unlabeled datasets has emerged as a dominant learning paradigm across multiple fields. This paradigm aligns well with the nature of medical imaging data, which has led to significant research efforts in applying self-supervised learning methods to this domain. However, many of these approaches fail to fully consider the unique characteristics of medical imaging, particularly the critical role that texture information plays in the diagnosis of thorax diseases. To address this gap, we propose a novel texture-aware self-supervised learning framework that leverages the Gray-Level Co-occurrence Matrix (GLCM) as an auxiliary signal to strengthen the model’s capacity to extract disease-relevant texture features. Additionally, we integrate curriculum learning into our approach, which gradually emphasizes texture information throughout the training process. This method enables the model to more effectively capture the inherent characteristics of medical imaging data. Our qualitative and quantitative experimental results show that our approach surpasses the current state-of-the-art methods on both the NIH CXR and Stanford CheXpert datasets.

Matching molecular analogues is a computational chemistry and bioinformatics research issue which is used to identify molecules that are structurally or functionally similar to a target molecule. Recent studies on matching analogous molecules have predominantly concentrated on enhancing effectiveness, often sidelining computational efficiency, particularly in contexts of low computational resources. This oversight poses challenges in many real applications (e.g., drug discovery, catalyst generation and so forth). To tackle this issue, we propose a general strategy named MapLE, aiming to promptly match analogous molecules with low computational resources by multi-metrics evaluation. Experimental evaluation conducted on a public biomolecular dataset validates the excellent and efficient performance of the proposed strategy.

Summary The dense and toxic blooms formed by cyanobacteria in aquatic environments pose significant threats to public health and aquatic ecosystems. Timely monitoring and prevention of cyanobacterial blooms in freshwater bodies are thus imperative. Although object detection methods have been applied in the field of algae identification, existing research faces several challenges. A primary issue is the overly idealistic setting of training sets, which are disconnected from the actual water quality environments, impeding practical algae identification and water quality monitoring. In this paper, we collect 2024 microscopic images of algae from a reservoir in Southern China, forming a comprehensive and diverse dataset for object detection. Addressing the aforementioned challenges, we propose an attention‐based strategy for the detection of tiny algal objects, which effectively samples and extracts features from algal targets. Our model uniquely leverages the morphable characteristics of algae, enhancing the accuracy and efficiency of identification. The training and improvement results of this model presented in our study are expected to aid in establishing a future system for real‐time algae monitoring and water quality assessment.

Gathering information from multi-perspective graphs is an essential issue for many applications especially for proteinligand binding affinity prediction. Most of traditional approaches obtained such information individually with low interpretability. In this paper, we harness the rich information from multi-perspective graphs with a general model, which abstractly represents protein-ligand complexes with better interpretability while achieving excellent predictive performance. In addition, we specially analyze the protein-ligand binding affinity problem, taking into account the heterogeneity of proteins and ligands. Experimental evaluations demonstrate the effectiveness of our data representation strategy on public datasets by fusing information from different perspectives.

Existing works have proved that using law articles as external knowledge can improve the performance of the Legal Judgment Prediction.However, they do not fully use law article information and most of the current work is only for single label samples.In this paper, we propose a Law Article Element-aware Multi-representation Model (LEMM), which can make full use of law article information and can be used for multi-label samples.The model uses the labeled elements of law articles to extract fact description features from multiple angles.It generates multiple representations of a fact for classification.Every label has a law-aware fact representation to encode more information.To capture the dependencies between law articles, the model also introduces a self-attention mechanism between multiple representations.Compared with baseline models like TopJudge, this model improves the accuracy of 5.84%, the macro F1 of 6.42%, and the micro F1 of 4.28%.

Enabling a mechanism to understand a temporal story and predict its ending is an interesting issue that has attracted considerable attention, as in case of the ROC Story Cloze Task (SCT). In this paper, we develop a multi-attention-based neural network (MANN) with well-designed optimizations, like Highway Network, and concatenated features with embedding representations into the hierarchical neural network model. Considering the particulars of the specific task, we thoughtfully extend MANN with external knowledge resources, exceeding state-of-the-art results obviously. Furthermore, we develop a thorough understanding of our model through a careful hand analysis on a subset of the stories. We identify what traits of MANN contribute to its outperformance and how external knowledge is obtained in such an ending prediction task.

To alleviate the error propagation in the traditional pipelined models for Abstract Meaning Representation (AMR) parsing, we formulate AMR parsing as a joint task that performs the two subtasks: concept identification and relation identification simultaneously.To this end, we first develop a novel componentwise beam search algorithm for relation identification in an incremental fashion, and then incorporate the decoder into a unified framework based on multiple-beam search, which allows for the bi-directional information flow between the two subtasks in a single incremental model.Experiments on the public datasets demonstrate that our joint model significantly outperforms the previous pipelined counterparts, and also achieves better or comparable performance than other approaches to AMR parsing, without utilizing external semantic resources.

Collaborative tagging services provided by various social web sites become popular means to mark web resources for different purposes such as categorization, expression of a preference and so on. However, the tags are of syntactic nature, in a free style and do not reflect semantics, resulting in the problems of redundancy, ambiguity and less semantics. Current tag-based recommender systems mainly take the explicit structural information among users, resources and tags into consideration, while neglecting the important implicit semantic relationships hidden in tagging data. In this study, we propose a Semantic Enhancement Recommendation strategy (SemRec), based on both structural information and semantic information through a unified fusion model. Extensive experiments conducted on two real datasets demonstarte the effectiveness of our approaches.