Xinyu Tang, Yuliang Zhan, Zhixun Li, Xin Zhao, Zhenduo Zhang, Zujie Wen, Zhiqiang Zhang, Jun Zhou. Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2026.

The advancement of single-cell RNA sequencing (scRNA-seq) technology has enabled researchers to capture cellular heterogeneity at the individual cell level, driving progress in diverse fields such as developmental biology, immunology, and cancer research. Accurate cell clustering is a crucial step for researchers utilizing scRNA-seq data; however, inherent characteristics like high dimensionality and sparsity pose significant challenges to obtaining precise clustering results. To achieve accurate clustering, this paper proposes a novel approach that integrates dimensionality reduction, self-representation matrix construction, and the clustering process into an end-to-end model termed LDSLRR (Low-Dimensional Sparse and Low-Rank Representation). Specifically, the original gene expression matrix first undergoes dimensionality reduction via projection. Subsequently, low-rank representation combined with a sparsity constraint facilitates the learning of the self-representation matrix. Finally, the cluster assignment matrix is acquired using graph-regularized non-negative matrix factorization (NMF). These three modules are simultaneously optimized, enhancing the accuracy of the clustering results. Comparative experiments against multiple state-of-the-art clustering methods on various scRNA-seq datasets demonstrate the superiority of the proposed LDSLRR method.

Multivariate time series forecasting is a crucial task that predicts the future states based on historical inputs. Although current deep learning-based methods have made significant advancements, they still face the criticism of lacking interpretability. The rise of the Kolmogorov-Arnold Network (KAN) provides a new perspective to implement an efficient and interpretable deep learning-based method for forecasting time series. However, we find there are two main challenges in the application of KAN in time series forecasting: how to select the appropriate one from various KAN variants and how to train the deep KAN-based network. To this end, we propose the multi-layer mixture-of-KAN network, which achieves excellent performance while retaining KAN's ability to be transformed into a combination of symbolic functions. The core module is the mixture-of-KAN layer, which uses a mixture-of-experts structure to assign variables to best-matched KAN experts. Then, we analyze the shortcomings of parameter initialization in the original KAN and provide an effective initialization method to alleviate training instability. Extensive experimental results demonstrate that our proposed method is effective in multivariate time series forecasting. Codes are released in https://github.com/2448845600/EasyTSF.

Abstract Direct kernel fuzzing is a targeted approach that focuses on specific areas of the kernel, effectively addressing the challenges of frequent updates and the inherent complexity of operating systems, which are critical infrastructure. This paper introduces SyzAgent, a framework integrating LLMs with the state-of-the-art kernel fuzzer Syzkaller, where the LLMs are used to guide the mutation and generation of test cases in real-time. We present preliminary results demonstrating that this method is effective on around 67% cases in our benchmark during the experiment.

Large reasoning models (LRMs) have demonstrated strong performance on complex reasoning tasks, but often suffer from overthinking, generating redundant content regardless of task difficulty. Inspired by the dual process theory in cognitive science, we propose Adaptive Cognition Policy Optimization (ACPO), a reinforcement learning framework that enables LRMs to achieve efficient reasoning through adaptive cognitive allocation and dynamic system switch. ACPO incorporates two key components: (1) introducing system-aware reasoning tokens to explicitly represent the thinking modes thereby making the model's cognitive process transparent, and (2) integrating online difficulty estimation and token length budget to guide adaptive system switch and reasoning during reinforcement learning. To this end, we propose a two-stage training strategy. The first stage begins with supervised fine-tuning to cold start the model, enabling it to generate reasoning paths with explicit thinking modes. In the second stage, we apply ACPO to further enhance adaptive system switch for difficulty-aware reasoning. Experimental results demonstrate that ACPO effectively reduces redundant reasoning while adaptively adjusting cognitive allocation based on task complexity, achieving efficient hybrid reasoning.

The advent and development of single-cell RNA sequencing (scRNA-seq) have provided new avenues for exploring cellular heterogeneity. Although many researchers have designed and developed efficient models to address cell heterogeneity and diversity by clustering cells into several groups, the performance of these methods may need improvement due to the characteristics of scRNA-seq data, such as high dimensionality, sparsity, and high dropout rates. In this paper, we propose a new method that couples low-rank representation (LRR) and concept factorization (CF) to learn a better clustering assignment matrix from both global and local perspectives, named SLRRGCF. Specifically, the LRR with similarity constraints based on tired random walk (TRW) can reduce the dimensionality of high-dimensional data while capturing more comprehensive global structure. At the same time, hypergraph regularization and CF are utilized to capture the local structure of the data further and directly obtain the clustering assignment matrix. We evaluated the performance of SLRRGCF on several real datasets, and comparisons with other competitive methods validated the effectiveness of our approach.

With the development of deep learning, traffic forecasting technology has made significant progress and is being applied in many practical scenarios. However, various events held in cities, such as sporting events, exhibitions, concerts, etc., have a significant impact on traffic patterns of surrounding areas, causing current advanced prediction models to fail in this case. In this paper, to broaden the applicable scenarios of traffic forecasting, we focus on modeling the impact of events on traffic patterns and propose an event traffic forecasting problem with multimodal inputs. We outline the main challenges of this problem: diversity and sparsity of events, as well as insufficient data. To address these issues, we first use textual modal data containing rich semantics to describe the diverse characteristics of events. Then, we propose a simple yet effective multi-modal event traffic forecasting model that uses pre-trained text and traffic encoders to extract the embeddings and fuses the two embeddings for prediction. Encoders pre-trained on large-scale data have powerful generalization abilities to cope with the challenge of sparse data. Next, we design an efficient large language model-based event description text generation pipeline to build multi-modal event traffic forecasting datasets, ShenzhenCEC and SuzhouIEC. Experiments on two real-world datasets show that our method achieves state-of-the-art performance compared with eight baselines, reducing mean absolute error during the event peak period by 4.26%. Code is available at: https://github.com/2448845600/EventTrafficForecasting.

Traffic forecasting is a critical task in intelligent transportation systems. In recent years, lots of methods have been proposed and achieved significant progress in modeling highly nonlinear and complex spatiotemporal pattern for traffic forecasting. However, most methods neglect the specific internal and external factors of the traffic system, such as the road connections, buildings surrounding each place, transfer stations, etc. The main challenges of utilizing the diverse knowledge from internal and external factors are to represent and fuse the impact of various factors. Few works use distance-based adjacency matrices to represent factors and the element-wise multiplication to fuse them, which may lead to even worse performances. In this paper, we propose to utilize knowledge graph to represent traffic system factors and design a novel neural network to exploit them for traffic foresting. First, we model the relations between factors and traffic conditions from the perspective of the knowledge graph and express them as unified triplets. Then, we generate multi-hop path features with embeddings learned from the knowledge representation model and multi-hop paths searched from the graph structure. Next, we present a knowledge-based multi-hop network (KMHNet) that uses an attention-based module to learn the correlation from multi-hop path features. Finally, to evaluate the performance of the proposed method, we build two real-world datasets both containing a traffic condition sub-dataset and a traffic knowledge graph. Experiments on two datasets demonstrate that our proposed KMHNet outperforms eight well-known methods. The code is publicly available at https://github.com/2448845600/KMHNet.

Learning an autonomous highlight video detector with good transferability across video categories, called Cross-Category Video Highlight Detection(CC-VHD), is crucial for the practical application on video-based media platforms. To tackle this problem, we first propose a framework that treats the CC-VHD as learning category-independent highlight feature representation. Under this framework, we propose a novel module, named Multi-task Feature Decomposition Branch which jointly conducts label prediction, cyclic feature reconstruction, and adversarial feature reconstruction to decompose the video features into two independent components: highlight-related component and category-related component. Besides, we propose to align the visual and audio modalities to one aligned feature space before conducting modality fusion, which has not been considered in previous works. Finally, the extensive experimental results on three challenging public benchmarks validate the efficacy of our paradigm and the superiority over the existing state-of-the-art approaches to video highlight detection.

Pose-invariant face recognition is a practically useful but challenging task. This paper introduces a novel method to learn pose-invariant feature representation without normalizing profile faces to frontal ones or learning disentangled features. We first design a novel strategy to learn pose-invariant feature embeddings by distilling the angular knowledge of frontal faces extracted by teacher network to student network, which enables the handling of faces with large pose variations. In this way, the features of faces across variant poses can cluster compactly for the same person to create a pose-invariant face representation. Secondly, we propose a Pose-Adaptive Angular Distillation loss to mitigate the negative effect of uneven distribution of face poses in the training dataset to pay more attention to the samples with large pose variations. Extensive experiments on two challenging benchmarks (IJB-A and CFP-FP) show that our approach consistently outperforms the existing methods.

Although face recognition has been widely applied in many areas, pose-robust face recognition is still a challenging topic due to the large pose variations in real scenes. In this paper, we propose to learn the pose-robust face representation by normalizing the profile face in feature level directly and jointly considering both intra-class compactness and inter-class separability. Our approach minimizes the angular distance between the profile face and the positive frontal anchor. And it maximizes the angular distance between the profile face and the negative frontal anchor simultaneously. Furthermore, we modify the Triplet loss and derive the Triplet Angular loss to guarantee the intra-class compactness and the inter-class separability in angular space. In this way, the faces under varying poses can cluster compactly to create a pose-robust feature representation. Extensive experiments on two challenging benchmarks (CFP-FP and IJB-A) illustrate that our approach achieves a competitive performance in the field of pose-robust face recognition.

By offloading computationally intensive tasks to the edge cloud provided by the cellular base stations, the mobile edge computing (MEC) technique has the potential to realize the critical millisecond-scale latency requirement of next generation mobile services. In this paper, we investigate the joint subcarrier and power allocation problem in an orthogonal frequency division multiple access (OFDMA) based MEC system to minimize the maximal delay of each mobile device. The partial data offloading scenario is considered where mobile data can be computed at both local devices and the edge cloud. Since the problem is a combinatorial optimization one, we first propose a lower-bound algorithm by relaxing the channel allocation indicators into continuous variables. Then, to find a low-complexity feasible solution, we further develop a heuristic algorithm which separates subcarrier assignment and power allocation. The performances of our proposed algorithms are finally tested by extensive numerical simulations.