Multi-hop question answering (MHQA) enables accurate answers to complex queries by retrieving and reasoning over evidence dispersed across multiple documents. Existing MHQA approaches mainly rely on iterative retrieval-augmented generation, which suffer from the following two major issues. On one hand, existing methods prematurely commit to surface-level entities rather than underlying reasoning structures, making question decomposition highly vulnerable to lexical ambiguity. On the other hand, existing methods overlook the logical dependencies among reasoning steps, resulting in uncoordinated execution. To address these issues, we propose STRIDE, a framework that separates strategic planning, dynamic control, and grounded execution. At its core, a Meta-Planner first constructs an entity-agnostic reasoning skeleton to capture the abstract logic of the query, thereby deferring entity grounding until after the reasoning structure is established, which mitigates disambiguation errors caused by premature lexical commitment. A Supervisor then orchestrates sub-question execution in a dependency-aware manner, enabling efficient parallelization where possible and sequential coordination when necessary. By dynamically deciding whether to retrieve new evidence or infer from existing facts, it avoids redundant queries and error propagation, while fusing cross-branch information and reformulating failed queries to enhance robustness. Grounded fact extraction and logical inference are delegated to specialized execution modules, ensuring faithfulness through explicit separation of retrieval and reasoning. While STRIDE is compatible with any large language models (LLMs), off-the-shelf open-source LLMs underperform closed-source counterparts in its structured reasoning pipeline. To close this gap, we further propose STRIDE-FT, a modular fine-tuning framework that uses self-generated execution trajectories from STRIDE, requiring neither human annotations nor stronger teacher models. Experiments show that STRIDE achieves robust and accurate reasoning on MHQA benchmarks, while STRIDE-FT effectively enhances open-source LLMs.

Lifelong model editing aims to dynamically adjust a model's output concerning specific facts, knowledge items, or behaviors, enabling the model to adapt to the evolving demands of realworld applications.While some retrieval-based methods have demonstrated potential in lifelong editing scenarios by storing edited knowledge in external memory, they often suffer from limitations in usability, such as requiring additional training corpora or lacking support for reversible and detachable edits.To address these issues, we propose a plug-and-play method for knowledge retrieval and injection, i.e., Layer-Level Prompting (LLP), which enables seamless and efficient lifelong model editing.In our LLP framework, the reasoning process of LLMs is divided into two stages, respectively, knowledge retrieval (Thinking) and knowledge injection (Recalling).Specifically, the knowledge retrieval process is performed in the early layers of the model, using layer outputs as thinking clues.And access the updated knowledge from memory in the subsequent layer to complete the knowledge injection process.Experimental results demonstrate that our method consistently outperforms existing techniques on lifelong model editing tasks, achieving superior performance on question answering and hallucination benchmarks across different LLMs.

Multivariate time series (MTS) classification has attracted increasing attention across various domains. Existing methods either decompose MTS into separate univariate series, ignoring inter-variable dependencies, or jointly model all variables, which may lead to over-smoothing and loss of semantic structure. These limitations become particularly pronounced when dealing with complex and heterogeneous variable types. To address these challenges, we propose SwinGroupNet (SGN), which explores a novel perspective for constructing variable interaction and temporal dependency. Specifically, SGN processes multi-scale time series using (1) Variable Group Embedding (VGE), which partitions variables into groups and performs independent group-wise embedding; (2) Multi-Scale Group Window Mixing (MGWM), which reconstructs variable interactions by modeling both intra-group and inter-group dependencies while extracting multi-scale temporal features; and (3) Periodic Window Shifting and Merging (PWSM), which exploits inherent periodic patterns to enable hierarchical temporal interaction and feature aggregation. Extensive experiments on diverse benchmark datasets from multiple domains demonstrate that SGN consistently achieves state-of-the-art performance, with an average improvement of 4.2% over existing methods. We release the source code at https://anonymous.4open.science/r/SGN.

Abstract Monads prove to be useful mathematical tools in theoretical computer science, notably in denoting different effects of programming languages. In this paper, we investigate a type of monads which arise naturally from Keimel and Lawson’s $\mathbf{K}$ -ification. A subcategory of $\mathbf{TOP}_{\mathbf{0}}$ is called of type $\mathrm{K}^{*}$ if it consists of monotone convergence spaces and is of type $\mathrm K$ in the sense of Keimel and Lawson. Each such category induces a canonical monad $\mathcal K$ on the category $\mathbf{DCPO}$ of dcpos and Scott-continuous maps, which is called the order- $\mathbf{K}$ -ification monad in this paper. First, for each category of type $\mathrm{K}^{*}$ , we characterize the algebras of the corresponding monad $\mathcal K$ as k-complete posets and algebraic homomorphisms as k-continuous maps, from which we obtain that the order- $\mathbf{K}$ -ification monad gives the free k-complete poset construction over the category $\mathbf{POS}_{\mathbf{d}}$ of posets and Scott-continuous maps. In addition, we show that all k-complete posets and Scott-continuous maps form a Cartesian closed category. Moreover, we consider the strongness of the order-K-ification monad and conclude with the fact that each order-K-ification monad is always commutative.

For the 6th generation (6G) communications, the interference exploitation generated from a growing number of intelligent factors with ultra scale is great challenge for the low complexity detection algorithms in the uplink multiuser massive multiple input and multiple output (MIMO) systems, especially for detecting high order quadrature amplitude modulation (QAM) signals. The deep learning technology is one of the key technical solutions for 6G. In this paper, a deep neural network based semidefinite relaxation (DNNSR) detection algorithm is proposed on the basis of the graphical detection model for uplink multiuser massive MIMO systems. Compared with the counterpart detection algorithms, along with a low polynomial average per symbol computational complexity, the proposed DNNSR detection algorithm requires lower average received signal to noise ratios to obtain better bit error rate performance as well as achieve theoretical spectral efficiency for graphical high order QAM signals in the uplink multiuser massive MIMO systems.

This study deals with the direction‐of‐arrival (DOA) estimation problem for hybrid completely polarised (CP) and partially polarised (PP) source signals using arbitrary polarimetric antenna arrays. An oblique projection‐based polarisation insensitive direction estimation (OPPIDE) algorithm is proposed by exploiting the spatial‐sparsity property of the sources. The OP technique is utilised to provide spatial filters, which are insensitive to the state of polarisation of signals, so that the potential source signals in the spatial domain can be separated later. The DOA estimation is finally implemented by identifying the sources’ spatially sparse structure with the separated signals. Theoretical analysis indicates that the OPPIDE is applicable to any hybrid CP and PP signals, and is independent of special polarimetric array configurations. The effectiveness and superiority of the proposed OPPIDE are substantiated through making performance comparison with the present counterpart algorithms.

Under no-reason return policy in an online setting, online retailers must determine whether to offer free shipping services when delivering products to consumers, and how to charge consumers for returns. To address such decision-making challenges, we develop a theoretic model to derive the optimal shipping strategy and return service charge (RSC) for an online retailer under two decision scenarios, in which decisions on shipping strategy and RSC are made either jointly or separately. We find that the retailer is better off in joint decision scenario than in separate decision scenario; a shipping free strategy is usually accompanied by a higher RSC, while a shipping fee strategy is typically accompanied by a lower RSC. We also find that, market parameters (e.g., the base return quantity, product price, consumers' sensitivities of shipping fee on demand, RSC on demand and return quantity) have important effects on the retailer's decisions on shipping strategy and RSC. Our findings suggest that the retailer can benefit from taking positive actions toward influencing the market to determine the favorable shipping strategy and RSC. Furthermore, our results can provide theoretical explanations for widely used shipping strategies and RSCs within the context of no-reason return policies in online settings. In particular, our analytical results explain why some real-world online retailers offer both free shipping and free return services to certain consumers.