Kai Yin, Xiangjue Dong, Chengkai Liu, Allen Lin, Lingfeng Shi, Ali Mostafavi, James Caverlee. Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2026.

Abstract Let $\textbf{Lat}_{\lor }$ denote the category of lattices with morphisms preserving supremum of non-empty finite subsets. We prove the following: (1) There exists a distributive law of the ordered-ideal monad over the ordered-filter monad on $\textbf{Lat}_{\lor }$. (2) The category ${\textbf{Cnt}}^{\mathbb{co}}$ of cocontinuous lattices with morphisms preserving arbitrary non-empty sups and filtered infs is strictly monadic over $\textbf{Lat}_{\lor }$. (3) The category ${\textbf{MCnt}}^{\mathbb{co}}$ of meet-continuous and cocontinuous lattices with morphisms preserving arbitrary non-empty sups and filtered infs is isomorphic to a full subcategory of ${\textbf{Lat}_{\lor }}^{\mathbb{FilId}}$, consisting of $\mathbb{FilId}$-algebras and $\mathbb{FilId}$-homomorphisms on ${\textbf{Lat}_{\lor }}$. For bounded-complete posets, analogous Eilenberg–Moore algebras provide a categorical representation. Finally, using the concept of bounded down-sets, we derive an algebraic characterization of ${{\textbf{Cnt}}}^{\mathrm{co}}$.

As people benefit from the convenience of social networks, they are also exposed to security risks from malicious information. This article proposes a malicious and anti-malicious information propagation dynamics model based on higher-order diffusion networks to address these issues. Firstly, to tackle the problem of individuals' behavior being swayed by the emotional content of malicious information during propagation, we have established an emotional influence mechanism. At the same time, adding user preferences and external drivers proposes a way to calculate individual influence. Secondly, in response to the discrepancies in the propagation efficiency of malicious information among varying users, we establish a higher-order diffusion network and quantify these differences through the stratification of users into three distinct layers of nodes. Additionally, considering the implicit relationship among potential user groups, we establish a group interaction mechanism by mining group attributes and refine the structure of the propagation network. Lastly, given the coexistence and opposition between malicious and anti-malicious information, dynamic game theory is applied to define a state transition equation incorporating anti-malicious information propagators. Consequently, we propose the SAIR model, a novel paradigm for understanding the propagation of malicious and anti-malicious information using higher-order diffusion networks.

Aiming to optimize marketing promotion, an information dissemination model integrating trust and user preference is developed. The objective is to capture users’ behavioral mechanisms and enhance marketing decision-making efficiency. To measure users’ trust in key opinion leaders, an Interval Type-2 Fuzzy Sets (IT2FSs) -based trust evaluation model is created, enabling effective trust assessment and stimulating purchasing behavior. Regarding the dynamic nature of user preferences, a Hidden Markov Model (HMM) -based prediction algorithm is proposed to track interest changes and forecast repurchase behavior. Considering rational and irrational user behaviors in marketing information dissemination, two new states, repurchase state P and hesitant purchasing state H are introduced based on the Susceptible-Infected-Recovered (SIR) model. Then, the SIRPH social platform information dissemination model is constructed, achieving accurate prediction and enhancement of marketing information dissemination. Experimental results indicate that the SIRPH model reduces the peak purchasing users by 15–25%, extends topic lifecycles by 20–30%, and improves information spreading accuracy, demonstrating the effectiveness of trust and preference integration.

Generative models have shown great promise in collaborative filtering by capturing the underlying distribution of user interests and preferences. However, existing approaches struggle with inaccurate posterior approximations and misalignment with the discrete nature of recommendation data, limiting their expressiveness and real-world performance. To address these limitations, we propose FlowCF, a novel flow-based recommendation system leveraging flow matching for collaborative filtering. We tailor flow matching to the unique challenges in recommendation through two key innovations: (1) a behavior-guided prior that aligns with user behavior patterns to handle the sparse and heterogeneous user-item interactions, and (2) a discrete flow framework to preserve the binary nature of implicit feedback while maintaining the benefits of flow matching, such as stable training and efficient inference. Extensive experiments demonstrate that FlowCF achieves state-of-the-art recommendation accuracy across various datasets with the fastest inference speed, making it a compelling approach for real-world recommender systems. The code is available at https://github.com/chengkai-liu/FlowCF.

Sequential recommender systems aims to predict the users' next interaction through user behavior modeling with various operators like RNNs and attentions. However, existing models generally fail to achieve the three golden principles for sequential recommendation simultaneously, i.e., training efficiency, low-cost inference, and strong performance. To this end, we propose RecBLR, an Efficient Sequential Recommendation Model based on Behavior-Dependent Linear Recurrent Units to accomplish the impossible triangle of the three principles. By incorporating gating mechanisms and behavior-dependent designs into linear recurrent units, our model significantly enhances user behavior modeling and recommendation performance. Furthermore, we unlock the parallelizable training as well as inference efficiency for our model by designing a hardware-aware scanning acceleration algorithm with a customized CUDA kernel. Extensive experiments on real-world datasets with varying lengths of user behavior sequences demonstrate RecBLR's remarkable effectiveness in simultaneously achieving all three golden principles - strong recommendation performance, training efficiency, and low-cost inference, while exhibiting excellent scalability to datasets with long user interaction histories.

We consider the problem of generating periodic materials with deep models. While symmetry-aware molecule generation has been studied extensively, periodic materials possess different symmetries, which have not been completely captured by existing methods. In this work, we propose SyMat, a novel material generation approach that can capture physical symmetries of periodic material structures. SyMat generates atom types and lattices of materials through generating atom type sets, lattice lengths and lattice angles with a variational auto-encoder model. In addition, SyMat employs a score-based diffusion model to generate atom coordinates of materials, in which a novel symmetry-aware probabilistic model is used in the coordinate diffusion process. We show that SyMat is theoretically invariant to all symmetry transformations on materials and demonstrate that SyMat achieves promising performance on random generation and property optimization tasks. Our code is publicly available as part of the AIRS library (https://github.com/divelab/AIRS).

In most real-world recommender systems, users interact with items in a sequential and multi-behavioral manner. Exploring the fine-grained relationship of items behind the users' multi-behavior interactions is critical in improving the performance of recommender systems. Despite the great successes, existing methods seem to have limitations on modelling heterogeneous item-level multi-behavior dependencies, capturing diverse multi-behavior sequential dynamics, or alleviating data sparsity problems. In this paper, we show it is possible to derive a framework to address all the above three limitations. The proposed framework MB-STR, a Multi-Behavior Sequential Transformer Recommender, is equipped with the multi-behavior transformer layer (MB-Trans), the multi-behavior sequential pattern generator (MB-SPG) and the behavior-aware prediction module (BA-Pred). Compared with a typical transformer, we design MB-Trans to capture multi-behavior heterogeneous dependencies as well as behavior-specific semantics, propose MB-SPG to encode the diverse sequential patterns among multiple behaviors, and incorporate BA-Pred to better leverage multi-behavior supervision. Comprehensive experiments on three real-world datasets show the effectiveness of MB-STR by significantly boosting the recommendation performance compared with various competitive baselines. Further ablation studies demonstrate the superiority of different modules of MB-STR.

Fault localization is one of the most important issues for the MMC consisting of large number of switches. This paper proposes a statistical multi-faults localization strategy for the MMC, where a feature extraction algorithm of capacitor voltage variations is proposed to extract the features of the MMC based on the exponential smoothing relationship among the capacitor voltages. Based on the extracted features, faults in the MMC can be easily located with the Grubbs Criterion according to the Grubbs Criterion Table. The proposed Grubbs Criterion-based fault localization strategy can construct concise simple features samples for the MMC containing all exponential smoothing global capacitor voltage variables, and accordingly it can locate faults with short time for the MMC. In addition, it not only does not require a large number of training data samples, but also does not require the creation of complex mathematical models and manual setting of empirical thresholds. It is compatible with both single and multiple switches open-circuit failures. The results of the simulation confirm the effectiveness of the proposed strategy.