Generative inference with Large Language Models (LLMs) is the cornerstone of web-scale AI search, where queries are answered using vast, heterogeneous documents retrieved via Retrieval-Augmented Generation (RAG). This paradigm is critically bottlenecked by the cost of self-attention mechanism on long context. The sheer diversity of retrieved web content (multi-sourced, multi-lingual, multi-faceted) makes simple Key-Value (KV) cache optimizations with pre-fixed subsets ineffective, demanding a dynamic, content-aware approach. This challenge, however, introduces a classic chicken-and-egg problem: the model cannot foresee the necessary KV entries for attention without first inferring on the content, yet doing so on the full context is prohibitively expensive. This paper introduces P&F, a unified framework that resolves this dilemma through a core ''probe-and-fetch'' mechanism, which ingeniously integrates with speculative decoding -- an acceleration approach already adopted in web-scale AI search. The probe step repurposes the speculative draft model: while generating candidate tokens, it simultaneously probes the context to predict the most salient KV entries the large model will need for attention. The fetch step immediately acts on this prediction, asynchronously fetching these sparse entries from memory. This synergistic design piggybacks the probing step onto the drafting process, allowing the expensive gathering of a sparse KV cache to be fully masked. Crucially, this co-design breaks the sequential dependency bottleneck that cripples naive integrations of speculative decoding and prefetching due to synchronization issues. Extensive experiments show P&F significantly outperforms state-of-the-art methods in throughput and scalability, offering a practical, drop-in solution. Extensive offline evaluations across various settings and datasets demonstrate that P&F yields superior throughput and scalability compared to advanced baselines, while maintaining model quality across diverse models and scales. In online settings, P&F delivers substantial gains in throughput improvements while preserving response quality, making it well-suited for large-scale industrial deployment in real-time AI Search services.

With the rise of AI in education, knowledge tracing (KT) has become important for modeling students’ knowledge from interaction data. However, existing methods still face three major challenges, including limited modeling of personalized exercise–concept relations, low robustness to noisy interactions, and inefficient training due to suboptimal data selection. To address these issues, we propose R 2 GCurL, a novel KT framework with two key designs. First, we recast KT as a graph classification problem and construct dynamic graphs from student responses, enabling the model to capture structural relations between exercises and concepts for more personalized KT. Second, we introduce a data-centric curriculum learning strategy based on dynamic graph entropy. Under our definition, pairwise dynamic graph entropy measures graph-transition continuity, where larger values indicate stronger structural similarity. Its sequence-level aggregation is used to derive a structure-aware difficulty signal for sample scheduling. On top of this, an RL-based scheduler further adapts batch selection based on model feedback and is especially beneficial under noisier and more unstable training regimes. Theoretical analysis shows that R 2 GCurL has lower computational complexity than existing graph-based KT models. Extensive experiments on five real-world datasets confirm its effectiveness, robustness, and generalizability, including as a plug-and-play enhancement for sequence-based KT models.

Graph contrastive learning has emerged as a powerful technique for learning graph representations that are robust and discriminative. However, traditional approaches often neglect the critical role of subgraph structures, particularly the intra-subgraph characteristics and inter-subgraph relationships, which are crucial for generating informative and diverse contrastive pairs. These subgraph features are crucial as they vary significantly across different graph types, such as social networks where they represent communities, and biochemical networks where they symbolize molecular interactions. To address this issue, our work proposes a novel subgraph-oriented learnable augmentation method for graph contrastive learning, termed SOLA-GCL, that centers around subgraphs, taking full advantage of the subgraph information for data augmentation. Specifically, SOLA-GCL initially partitions a graph into multiple densely connected subgraphs based on their intrinsic properties. To preserve and enhance the unique characteristics inherent to subgraphs, a graph view generator optimizes augmentation strategies for each subgraph, thereby generating tailored views for graph contrastive learning. This generator uses a combination of intra-subgraph and inter-subgraph augmentation strategies, including node dropping, feature masking, intra-edge perturbation, inter-edge perturbation, and subgraph swapping. Extensive experiments have been conducted on various graph learning applications, ranging from social networks to molecules, under semi-supervised learning, unsupervised learning, and transfer learning settings to demonstrate the superiority of our proposed approach.

While representation learning has been used to boost the performance of Learning-to-Rank (LTR) models through distilling key features for webpage ranking, the weak supervision signals extracted from users' sparse click-through data lead to inadequate representation of query-webpage pairs for ranking score prediction. Recent studies in generative LTR pre-training demonstrate the feasibility of incorporating reconstruction loss for enhanced ranking score prediction. However, LTR is afterall a regression task and it might be reasonable to find an alternate route that pre-trains LTR models with discriminative losses. Following the success of Electra in representation learning for natural language processing (NLP), this work proposes RankElectra that pre-trains the LTR model as a discriminator module inside a generative learning framework. Specifically, RankElectra first structures sparsely-annotated query-webpage pairs into a bipartite graph, with query and webpage feature vectors as node types and ranking scores as the connecting edges, and then leverages positive and negative extension strategies to densify the graph by link predictions. Later, this work proposes a novel Electra module that pre-trains the LTR model as a discriminator module for node reconstruction tasks, where node features of selected edges would be randomly masked and reconstructed by a generator, and the discriminator learns to classify whether the reconstructed features are the original or replaced as well as perform correct ranking. Finally, the pre-trained discriminator module, rather than the generator, would be fine-tuned on the labeled graph. We carried out extensive offline and online evaluations using the real-world web traffic of Baidu search engine. The results show that RankElectra could significantly boost the ranking performance of Baidu Search compared with numbers of competitor systems.

EEG-based emotion recognition plays an important role in developing adaptive brain-computer communication systems, yet faces two fundamental challenges in practical implementations: (1) effective integration of non-stationary spatial-temporal neural patterns, (2) robust adaptation to dynamic emotional intensity variations in real-world scenarios. This paper proposes STT-CL, a novel framework integrating spatial-temporal transformers with curriculum learning. Our method introduces two core components: a spatial encoder that models inter-channel relationships and a temporal encoder that captures multi-scale dependencies through windowed attention mechanisms, enabling simultaneous extraction of spatial correlations and temporal dynamics from EEG signals. Complementing this architecture, an intensity-aware curriculum learning strategy progressively guides training from high-intensity to low-intensity emotional states through dynamic sample scheduling based on a dual difficulty assessment. Comprehensive experiments on three benchmark datasets demonstrate state-of-the-art performance across various emotional intensity levels, with ablation studies confirming the necessity of both architectural components and the curriculum learning mechanism.

Sequential recommendation systems aim to predict the future behaviors of users based on their historical interactions. Despite the success of neural architectures like Transformer and Graph Neural Networks, these models often struggle with the inherent challenge of sparse data in accurately predicting future user behaviors. To alleviate the data sparsity problem, some methods leverage the contrastive learning to generate contrastive views, assuming the items appear discretely at the same time intervals and focusing on the sequence order. However, these approaches neglect the crucial temporal-aware collaborative patterns hidden within the user-item interactions, leading to a limited variety of contrastive pairs and less informative embeddings. The proposed framework,Temporal-awaregraph contrastive learning with theoretical guarantees for sequentialRecommendation (TagRec), integrates temporal-aware collaborative patterns with adaptive data augmentation to generate more informative user and item representations. TagRec employs a temporal-aware graph neural network to embed the original graph, then generates augmented graphs through the addition of interactions via latent user interest mining, the dropping of redundant interaction edges, and the perturbation of temporal information. Theoretical guarantees are provided that these augmentations enhance the graph’s utility. Extensive experiments on real-world datasets demonstrate the superiority of the proposed approach over the state-of-the-art recommendation methods.

LLM has achieved impressive performance on multi-modal tasks, which have received everincreasing research attention.Recent research focuses on improving prediction performance and reliability (e.g., addressing the hallucination problem).They often prepend relevant external knowledge to the input text as an extra prompt.However, these methods would be affected by the noise in the knowledge and the context length limitation of LLM.In our work, we focus on making better use of external knowledge and propose a method to actively extract valuable information in the knowledge to produce the latent vector as a soft prompt, which is then fused with the image embedding to form a knowledge-enhanced context to instruct LLM.The experimental results on knowledge-based VQA benchmarks show that the proposed method enjoys better utilization of external knowledge and helps the model achieve better performance.

Graph neural networks (GNNs) have shown ad-vantages in graph-based analysis tasks. However, most existing methods have the homogeneity assumption and show poor performance on heterophilic graphs, where the linked nodes have dissimilar features and different class labels, and the semantically related nodes might be multi-hop away. To address this limitation, this paper presents GraphRARE, a general framework built upon node relative entropy and deep reinforcement learning, to strengthen the expressive capability of GNNs. An innovative node relative entropy, which considers node features and structural similarity, is used to measure mutual information between node pairs. In addition, to avoid the sub-optimal solutions caused by mixing useful information and noises of remote nodes, a deep reinforcement learning-based algorithm is developed to optimize the graph topology. This algorithm selects informative nodes and discards noisy nodes based on the defined node relative en-tropy. Extensive experiments are conducted on seven real-world datasets. The experimental results demonstrate the superiority of GraphRARE in node classification and its capability to optimize the original graph topology.

Knowledge tracing models based on deep learning can achieve impressive predictive performance by leveraging attention mechanisms. However, there still exist two challenges in attentive knowledge tracing (AKT): First, the mechanism of classical models of AKT demonstrates relatively low attention when processing exercise sequences with shifting knowledge concepts (KC), making it difficult to capture the comprehensive state of knowledge across sequences. Second, classical models do not consider stochastic behaviors, which negatively affects models of AKT in terms of capturing anomalous knowledge states. This article proposes a model of AKT, called Enhancing Locality for Attentive Knowledge Tracing (ELAKT), that is a variant of the deep KT model. The proposed model leverages the encoder module of the transformer to aggregate knowledge embedding generated by both exercises and responses over all timesteps. In addition, it uses causal convolutions to aggregate and smooth the states of local knowledge. The ELAKT model uses the states of comprehensive KCs to introduce a prediction correction module to forecast the future responses of students to deal with noise caused by stochastic behaviors. The results of experiments demonstrated that the ELAKT model consistently outperforms state-of-the-art baseline KT models.

Modeling and predicting the performance of students in collaborative learning paradigms is an important task. Most of the research presented in literature regarding collaborative learning focuses on the discussion forums and social learning networks. There are only a few works that investigate how students interact with each other in team projects and how such interactions affect their academic performance. In order to bridge this gap, we choose a software engineering course as the study subject. The students who participate in a software engineering course are required to team up and complete a software project together. In this work, we construct an interaction graph based on the activities of students grouped in various teams. Based on this student interaction graph, we present an extended graph transformer framework for collaborative learning (CLGT) for evaluating and predicting the performance of students. Moreover, the proposed CLGT contains an interpretation module that explains the prediction results and visualizes the student interaction patterns. The experimental results confirm that the proposed CLGT outperforms the baseline models in terms of performing predictions based on the real-world datasets. Moreover, the proposed CLGT differentiates the students with poor performance in the collaborative learning paradigm and gives teachers early warnings, so that appropriate assistance can be provided.