Shuyuan Zhao, Wei Chen, Weijie Zhang, Xinrui Hou, Junfeng Shen, Boyan Shi, Shengnan Guo, Youfang Lin, Huaiyu Wan. Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2026.

Reasoning future unknowable facts on temporal knowledge graphs (TKGs) is a challenging task, holding significant academic and practical values for various fields. Existing studies exploring explainable reasoning concentrate on modeling comprehensible temporal paths relevant to the query. Yet, these path-based methods primarily focus on local temporal paths appearing in recent times, failing to capture the complex temporal paths in TKG and resulting in the loss of longer historical relations related to the query. Motivated by the Dual Process Theory in cognitive science, we propose a Cognitive Temporal Knowledge Extrapolation framework (CognTKE), which introduces a novel temporal cognitive relation directed graph (TCR-Digraph) and performs interpretable global shallow reasoning and local deep reasoning over the TCR-Digraph. Specifically, the proposed TCR-Digraph is constituted by retrieving significant local and global historical temporal relation paths associated with the query. In addition, CognTKE presents the global shallow reasoner and the local deep reasoner to perform global one-hop temporal relation reasoning (System 1) and local complex multi-hop path reasoning (System 2) over the TCR-Digraph, respectively. The experimental results on four benchmark datasets demonstrate that CognTKE achieves significant improvement in accuracy compared to the state-of-the-art baselines and delivers excellent zero-shot reasoning ability.

The takeaway recommendation system aims to recommend users' future takeaway purchases based on their historical purchase behaviors, thereby improving user satisfaction and boosting merchant sales. Existing methods focus on incorporating auxiliary information or leveraging knowledge graphs to alleviate the sparsity issue of user purchase sequences. However, two main challenges limit the performance of these approaches: (1) capturing dynamic user preferences on complex geospatial information and (2) efficiently integrating spatial-temporal knowledge from both graphs and sequence data with low computational costs. In this paper, we propose a novel spatial-temporal knowledge distillation model for takeaway recommendation (STKDRec) based on the two-stage training process. Specifically, during the first pre-training stage, a spatial-temporal knowledge graph (STKG) encoder is trained to extract high-order spatial-temporal dependencies and collaborative associations from the STKG. During the second spatial-temporal knowledge distillation (STKD) stage, a spatial-temporal Transformer (ST-Transformer) is employed to comprehensively model dynamic user preferences on various types of fine-grained geospatial information from a sequential perspective. Furthermore, the STKD strategy is introduced to transfer graph-based spatial-temporal knowledge to the ST-Transformer, facilitating the adaptive fusion of rich knowledge derived from both the STKG and sequence data while reducing computational overhead. Extensive experiments on three real-world datasets show that STKDRec significantly outperforms the state-of-the-art baselines.

Recent Continual Learning (CL)-based Temporal Knowledge Graph Reasoning (TKGR) methods focus on significantly reducing computational cost and mitigating catastrophic forgetting caused by fine-tuning models with new data. However, existing CL-based TKGR methods still face two key limitations: (1) They usually one-sidedly reorganize individual historical facts, while overlooking the historical context essential for accurately understanding the historical semantics of these facts; (2) They preserve historical knowledge by simply replaying historical facts, while ignoring the potential conflicts between historical and emerging facts. In this paper, we propose a \textbf{D}eep \textbf{G}enerative \textbf{A}daptive \textbf{R}eplay (DGAR) method, which can generate and adaptively replay historical entity distribution representations from the whole historical context. To address the first challenge, historical context prompts as sampling units are built to preserve the whole historical context information. To overcome the second challenge, a pre-trained diffusion model is adopted to generate the historical distribution. During the generation process, the common features between the historical and current distributions are enhanced under the guidance of the TKGR model. In addition, a layer-by-layer adaptive replay mechanism is designed to effectively integrate historical and current distributions. Experimental results demonstrate that DGAR significantly outperforms baselines in reasoning and mitigating forgetting.

Recent studies have combined Mixture of Experts (MoE) and Parameter-Efficient Fine-tuning (PEFT) to fine-tune large language models (LLMs), holding excellent performance in multi-task scenarios while remaining resource-efficient. However, existing MoE approaches still exhibit the following limitations: (1) Current methods fail to consider that different LLM layers capture features at varying levels of granularity, leading to suboptimal performance. (2) Task-level routing methods lack generalizability to unseen tasks. (3) The uncertainty introduced by load imbalance loss undermines the effective specialization of the experts. To address these challenges, we propose HMoRA, a Hierarchical fine-tuning method that combines MoE and LoRA, employing hybrid routing that integrates token-level and task-level routing in a hierarchical manner. This hierarchical hybrid routing allows the model to more efficiently capture both fine-grained token information and broader task contexts. To improve the certainty of expert selection, a novel routing auxiliary loss is introduced. This auxiliary function also enhances the task router's ability to differentiate tasks and its generalization to unseen tasks. Additionally, several optional lightweight designs have been proposed to significantly reduce both the number of trainable parameters and computational costs. Experimental results demonstrate that HMoRA outperforms full fine-tuning across multiple NLP benchmarks, while fine-tuning only 3.9\% of the parameters. The code is available on: https://github.com/LiaoMengqi/HMoRA.

Next POI (Point-of-Interest) recommendation aims to forecast users’ future movements based on their historical check-in trajectories, holding significant value in location-based services. Existing methods address trajectory data sparsity by integrating rich auxiliary information or using spatial-temporal knowledge graphs (STKGs), showing promising results. Yet, they face two main challenges: i) Due to the difficulty of transforming structured trajectory data into trajectory text describing users’ spatial-temporal mobility, the powerful reasoning ability of pre-trained language models is rarely explored to enhance recommendation performance. ii) Methods based on STKG can introduce external knowledge inconsistent with user preferences, leading to the knowledge noise generated hampering the accuracy of recommendations. To this end, we propose a novel approach called STKG-PLM that integratesSTKGcontrastive learning andprompt pre-trainedlanguagemodel (PLM) to enhance the next POI recommendation. Specifically, we design a spatial-temporal trajectory prompt template that transforms structured trajectories into text corpus based on STKG, serving as the input of PLM to understand the movement pattern of users from coarse-grained and fine-grained perspectives. Additionally, we propose an STKG contrastive learning framework to mitigate the introduced knowledge noise. Extensive experiments on three real-world datasets demonstrate that STKG-PLM exhibits notable performance improvements over the state-of-the-art baseline methods.

Zero-shot stance detection on social media (ZSSD-SM) aims to distinguish the attitude in tweets towards an unseen target. Previous work capture latent variables between source and target domains to perform this task, but the lack of context knowledge hinders the detection performance. Recent studies have been devoted to obtaining the accurate representation of tweets by bringing additional facts from Knowledge Graph (KG), showing promising performance. However, these knowledge injection methods still suffer from two challenges: (i) The pipeline of knowledge injection causes error accumulation and (ii) irrelevant knowledge makes them fail to understand the semantics. In this paper, we propose a novel knowledge injection method for ZSSD-SM, which adopts two training stages, namely knowledge compression and task guidance, to flexibly inject knowledge into the pre-trained language model (PLM) and adaptively expand tweets context. Specifically, in the knowledge compression stage, the latent representation of KG is reconstructed by the triplet denoising task and compressed into external matrices; while in the task guidance stage, the frozen matrices are employed to guide the PLM to adaptively extract its own context-related knowledge, and then complete the fine-tuning of the ZSSD-SM task. Extensive experiments on multiple datasets show the effectiveness of our proposed method. The code is available at: https://github.com/ShuohaoLin/KPatch.

Temporal knowledge graphs (TKGs) have been identified as a promising approach to represent the dynamics of facts along the timeline. The extrapolation of TKG is to predict unknowable facts happening in the future, holding significant practical value across diverse fields. Most extrapolation studies in TKGs focus on modeling global historical fact repeating and cyclic patterns, as well as local historical adjacent fact evolution patterns, showing promising performance in predicting future un-known facts. Yet, existing methods still face two major challenges: (1) They usually neglect the importance of historical information in KG snapshots related to the queries when encoding the local and global historical information; (2) They exhibit weak anti-noise capabilities, which hinders their performance when the inputs are contaminated with noise. To this end, we propose a novel Local-global history-aware Contrastive Learning model (LogCL) for TKG reasoning, which adopts contrastive learning to better guide the fusion of local and global historical information and enhance the ability to resist interference. Specifically, for the first challenge, LogCL proposes an entity-aware attention mechanism applied to the local and global historical facts encoder, which captures the key historical information related to queries. For the latter issue, LogCL designs a local-global query contrast module, effectively improving the robustness of the model. The experimental results on four benchmark datasets demonstrate that LogCL delivers better and more robust performance than the state-of-the-art baselines. The code of LogCL is available at https://eithub.com/WeiChen3690/LoeCL.

Multi-modal data have recently been utilized to improve the performance of knowledge graph completion (KGC), attracting widespread research interest. However, they have been ignored in few-shot knowledge graph completion (FKGC), which aims to discover potential facts involving unseen relations that only appear in few-shot triples. The most relevant FKGC study simply concatenates various modal features, but the performance is still limited due to the following problems: (1) lack of exploiting significant multi-modal features in neighborhoods, and (2) ineffectively modeling inter-modal interactions in a few-shot setting. To tackle these problems, we propose a novel relational learning model entitled MMSN (Multi-Modal Siamese Network) for few-shot knowledge graph completion, which is composed of the following two primary modules: the Siamese multi-modal neighbor encoder (SMNE) and the meta-learning multi-modal knowledge representation decoder (MKRD). The module SMNE is developed to encode diverse modalities of neighbors by a Siamese attention network, fuse multi-modal information through a gating fusion network, and learn effective relational embeddings using an aggregator. The module MKRD is introduced to handle inter-modal interactions between multiple modalities and train the proposed model in a few-shot scenario. Extensive experiments demonstrate that our proposed model MMSN outperforms the state-of-the-art FKGC models, including uni-modal and multi-modal models, on two real-world few-shot multi-modal datasets.