Mixture-of-Experts (MoE) architectures have recently become a more prevalent choice for large language models (LLMs) than dense architectures due to their superior performance. However, billions of parameters bring MoE LLMs a huge cost for deployment and inference. To address these issues, knowledge distillation (KD) has become a widely adopted technique to compress LLMs. Existing KD methods for LLMs can be divided into dense-to-dense and moe-to-dense distillation. Dense-to-dense distillation transfers knowledge between single dense LLMs, while moe-to-dense distillation attempts to transfer knowledge between the MoE LLMs and the dense LLMs. However, the architectural mismatch prevents the student from fully absorbing knowledge when distilling MoE LLMs. To address this limitation, we investigate a new distillation setting, moe-to-moe, which aims to fully leverage expert knowledge of teachers and enable the student to absorb it more effectively. Compared to dense-to-dense and moe-to-dense, moe-to-moe suffers from two imbalance issues. First, expert-coverage deficiency reflects an imbalanced knowledge transfer of teacher experts: traditional distillation utilizes only the few experts activated by the teacher router. Second, routing imbalance appears when the student routing distribution drifts from the teacher, which makes it difficult for students to learn how to distribute different experts. To overcome these issues, we propose a novel distillation framework for moe-to-moe, Balanced Distillation (B-Distill), which equally spreads teacher expertise across student experts while regularizing the student router toward teacher-consistent balance. First, to mitigate expert-coverage deficiency, we introduce Monte Carlo exploration, which stochastically perturbs router probabilities so every teacher and student expert is sampled without enlarging the search space. Second, to correct routing imbalance and avert load collapse, we propose an entropy-aware router distillation mechanism that aligns the student router with the teacher while curbing over-concentration. Experiments show that B-Distill outperforms baselines by up to 6.6% in Rouge-L.

Mixture of experts (MoE) dynamically routes inputs to specialized expert networks to scale model capacity with low inference overhead. However, the excessive parameter growth in MoE models poses challenges in low-resource settings. To address these issues, MoE with parameter-efficient fine-tuning (PEFT) methods have emerged as a lightweight adaptation paradigm that distributes knowledge among experts via multiple LoRA blocks. Existing MoE-PEFT methods can be broadly categorized into External and Internal PEFT methods. External PEFT methods incorporate lightweight models into existing MoE architectures without modifying their routing, which limits the model’s parameter efficiency. To overcome these issues, Internal PEFT methods integrate MoE architectures into PEFT, enabling minimal parameter overhead. However, they still face two major challenges: (1) lack of expert functional differentiation, resulting in overlapping specialization across modules, and (2) absence of a structured attribution mechanism to guide expert selection based on semantic relevance. To alleviate these challenges, we propose TopicLoRA, a novel three-stage framework that leverages topic knowledge as semantic anchors to guide expert allocation. Specifically, (1) to address expert redundancy, we construct a topic-level prior graph using Graph Neural Network-enhanced representation learning over Big-Bench categories, enforcing structural separation among expert embeddings, and (2) to introduce semantic attribution, we design a dual-loss training mechanism that softly aligns input-query relevance with topic-guided routing distributions via KL divergence. Extensive experiments on representative datasets (e.g., MMLU, GSM8K, Flanv2) demonstrate that TopicLoRA outperforms state-of-the-art PEFT baselines by 2.40% on average in accuracy. Notably, the maximum improvement is 4.21%. Furthermore, ablation studies demonstrate that our framework's robustness to intricate topics and input sequence variations, which stems from the dual-loss training mechanism.

Large Language Models (LLMs) have demonstrated strong capabilities across diverse tasks under the example-driven learning paradigm. However, in high-stakes domains such as emergency response and industrial safety, historical incidents are scarce, confidential, or both, while concise rule books are abundant. We formalize this underexplored setting as rule knowledge-driven reasoning and ask: Can LLMs reason reliably when rules are plentiful but examples are nearly absent? To study this question, we introduce RULER, an automatic benchmark that generates 32K rigorously verified questions from 1K expert-curated emergency response rules to probe three core abilities: rule memorization, single-rule application, and multi-rule complex reasoning. RULER is further equipped with a hallucination-aware evaluation suite and novel relational metrics. A comprehensive empirical study of five representative LLMs and five enhancement strategies shows that, even when models achieve reliable performance on rule memorization and single-rule application, multi-rule complex reasoning plateaus at 5.4 on a 10-point scale. To address this limitation, we propose RAMPS, a Rule knowledge-Aware Monte Carlo Tree Search Process-reward Supervision framework. RAMPS injects rule knowledge priors into MCTS, distills 12K step-level traces without human annotation, and trains an advantage-based reward model that scores candidate reasoning paths during beam search inference. Experimental results show that RAMPS significantly improves multi-rule complex reasoning performance to 7.7.

Relation extraction (RE) identifies semantic relations between entities in text, with existing methods falling into two main paradigms: discriminative and generative.Discriminative models encode sentences and entities into relation representations and classify the most likely relation, whereas generative models directly produce relation labels through sequence generation.Although the latter have benefited from recent advances in large language models (LLMs), their performance remains limited by bottlenecks.In this work, we present the systematic investigation of how discriminative models can support generative RE.We propose the Discriminative-to-Generative (D2G) framework, which first leverages discriminative models to produce a top-k set of candidate relations, and then integrates this knowledge into generative models via in-context or prompt learning.Extensive experiments on five benchmarks demonstrate that D2G consistently achieves state-of-the-art performance, with notable gains on long-tailed relation classes.

Chuanxin Zhang, Jiajun Liu, Yao He, Wenjun Ke, Peng Wang, Yankun Le, Sirui Liu, Zhaoyu Yang. Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2026.

Unified Information Extraction (UIE) aims to handle heterogeneous IE tasks within a single framework, but existing methods often suffer from inconsistent schema representation, implicitly intermediate reasoning and fullparameter adaptation, which limit generalization, interpretability and parameter efficiency.To address these issues, we propose UC-UIE (Universal Capabilities-based Unified Information Extractor), a unified framework based on Large Language Model (LLM), which introduces a unified frame-and-slots schema for IE tasks and explicitly decomposes IE reasoning into three universal capabilities: judging, locating, and associating.Furthermore, UC-UIE adopts a Low-Rank Adaptation (LoRA) based hierarchical Mixture-of-Experts (MoE) adapter to fine-tune LLMs for IE tasks, which explicitly models these three capabilities in a task-driven way while ensuring parameter efficiency.With only 1.24% trainable parameters, UC-UIE outperforms full-parameter tuning methods, showing excellent parameter efficiency.Zero-shot evaluation reveals its strong generalization ability to unseen domains and schemas, benefiting from unified schema representation and explicit capability decomposition.Further experiments validate that the hierarchical MoE adapter learns capability specialization and composition, which enhances both UIE performance and interpretability.

Existing knowledge graphs (KGs) inevitably contain outdated or erroneous knowledge that needs to be removed from knowledge graph embedding (KGE) models. To address this challenge, knowledge unlearning can be applied to eliminate specific information while preserving the integrity of the remaining knowledge in KGs. Existing unlearning methods can generally be categorized into exact unlearning and approximate unlearning. However, exact unlearning requires high training costs, while approximate unlearning faces two issues when applied to KGs due to the inherent connectivity of triples: (1) It fails to fully remove targeted information, as forgetting triples can still be inferred from remaining ones. (2) It focuses on local data for specific removal, which weakens the remaining knowledge in the forgetting boundary. To address these issues, we propose GraphDPO, a novel approximate unlearning framework based on direct preference optimization (DPO). Firstly, to effectively remove forgetting triples, we reframe unlearning as a preference optimization problem, where the model is trained by DPO to prefer reconstructed alternatives over the original forgetting triples. This formulation penalizes reliance on forgettable knowledge, mitigating incomplete forgetting caused by KG connectivity. Moreover, we introduce an out-boundary sampling strategy to construct preference pairs with minimal semantic overlap, weakening the connection between forgetting and retained knowledge. Secondly, to preserve boundary knowledge, we introduce a boundary recall mechanism that replays and distills relevant information both within and across time steps. We construct eight unlearning datasets across four popular KGs with varying unlearning rates. Experiments show that GraphDPO outperforms state-of-the-art baselines by up to 10.1% in MRR_Avg and 14.0% in MRR_F1. Further analysis confirms that GraphDPO more effectively removes target knowledge while preserving surrounding context.

The rapid proliferation of documents has made document intelligence increasingly critical across various industries. In recent years, Large Language Models (LLMs) have dramatically transformed the field of document intelligence, allowing for more advanced and accurate document processing solutions. Despite these advancements, most existing surveys have failed to focus on these breakthroughs, instead concentrating on traditional methods and earlier machine learning techniques. This survey seeks to fill that gap by offering an in-depth analysis of approximately 300 papers published between 2021 and mid-2025, thus providing a comprehensive overview of the impact of LLMs in document intelligence. The key topics explored include Retrieval-Augmented Generation (RAG), long-context processing, and fine-tuning LLMs for document comprehension. Furthermore, the survey highlights essential datasets, practical applications, current challenges, and future research directions, offering critical insights for both researchers and industry practitioners looking to advance the field.

Topic modeling aims to discover the distribution of topics within a corpus. The advanced comprehension and generative capabilities of large language models (LLMs) have introduced new avenues for topic modeling, particularly by prompting LLMs to generate topics and refine them by merging similar ones. However, this approach necessitates that LLMs generate topics with consistent granularity, thus relying on the exceptional instruction-following capabilities of closed-source LLMs (such as GPT-4) or requiring additional training. Moreover, merging based only on topic words and neglecting the fine-grained semantics within documents might fail to fully uncover the underlying topic structure. In this work, we propose a semi-supervised topic modeling method, LiSA, that combines LLMs with clustering to improve topic generation and distribution. Specifically, we begin with prompting LLMs to generate a candidate topic word for each document, thereby constructing a topic-level semantic space. To further utilize the mutual complementarity between them, we first cluster documents and candidate topic words, and then establish a mapping from document to topic in the LLM-guided assignment stage. Subsequently, we introduce a collaborative enhancement strategy to align the two semantic spaces and establish a better topic distribution. Experimental results demonstrate that LiSA outperforms state-of-the-art methods that utilize GPT-4 on topic alignment, and exhibits competitive performance compared to Neural Topic Models on topic quality. The codes are available at https://github.com/ljh986/LiSA.

Ziyu Shang, Jianghan Liu, Zhizhao Luo, Peng Wang, Wenjun Ke, Jiajun Liu, Zijie Xu, Guozheng Li. Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2025.

Prompt optimization aims to construct effective prompts to fully leverage the capabilities of large language models (LLMs). The key challenge lies in analyzing the outputs of LLMs, identifying fine-grained errors as explicit evidence and guiding reverse prompt correction. However, existing dominant methods feed all evidence into LLMs, leading to excessive redundancy that misleads the models. Moreover, this redundant input could increase computational costs. To address the issue, we introduce a novel meta evidence-aware prompt optimization method (MEPO), which performs fine-grained analysis of the causal relationships within evidence using clustering and Bayesian inference. Initially, MEPO generates preliminary evidence to describe the drawbacks of the given prompt. Subsequently, the evidence is clustered and constructed into a directed acyclic graph (DAG) to produce meta evidence. Finally, we edit the initial prompt based on the obtained meta evidence and employ the beam search to select the best-performing prompt. Experimental results show that MEPO outperforms state-of-the-art baselines, achieving absolute F1 score improvements of 5.8% and 5.0% on two common-used Liar and Jailbreak datasets, respectively. The datasets and codes are available at https://anonymous.4open.science/r/MEPO-E1A0.

Knowledge graph reasoning (KGR) seeks to infer new factual triples from existing knowledge graphs (KGs). Recent methods have unified transductive and inductive reasoning by learning entity-independent representations through local neighboring structures. Nevertheless, these methods often encounter inefficiencies and rely on elaborate local structures without directly modeling the correlations between queries and various structures within KGs. In this paper, we propose a novel framework MulGA, which is designed to learn multi-granularity and adaptive embeddings for KGR. MulGA first employs connectivity subgraphs to uniformly and hierarchically represent query-related structures within KGs, such as triples, relation paths, and subgraphs, establishing the hierarchical relationship between structures at different granularities. Subsequently, we design a graph neural network-based multi-granularity embedding propagation module that unifies the message-passing process with the connectivity subgraph construction. This module obtains the query-related structural representations by all entities at multiple granularities, eliminating the need to explicitly extract any graph elements, thus addressing inefficiency issues. Moreover, we develop a structure-aware adaptive merging mechanism that assigns weights to different granularities and integrates them into cohesive subgraph-granularity representations for reasoning. The systematic experiments have been conducted on 15 benchmarks and MulGA achieves a significant improvement in MRR by an average of 0.5%-1.1% on transductive tasks and 0.2%-7.3% on inductive tasks than existing state-of-the-art methods. Moreover, MulGA exhibits faster convergence speed, smaller number of parameters, competitive inference time, and alleviates the over-smoothing prevalent in graph neural networks.

Traditional knowledge graph embedding (KGE) methods typically require preserving the entire knowledge graph (KG) with significant training costs when new knowledge emerges. To address this issue, the continual knowledge graph embedding (CKGE) task has been proposed to train the KGE model by learning emerging knowledge efficiently while simultaneously preserving decent old knowledge. However, the explicit graph structure in KGs, which is critical for the above goal, has been heavily ignored by existing CKGE methods. On the one hand, existing methods usually learn new triples in a random order, destroying the inner structure of new KGs. On the other hand, old triples are preserved with equal priority, failing to alleviate catastrophic forgetting effectively. In this paper, we propose a competitive method for CKGE based on incremental distillation (IncDE), which considers the full use of the explicit graph structure in KGs. First, to optimize the learning order, we introduce a hierarchical strategy, ranking new triples for layer-by-layer learning. By employing the inter- and intra-hierarchical orders together, new triples are grouped into layers based on the graph structure features. Secondly, to preserve the old knowledge effectively, we devise a novel incremental distillation mechanism, which facilitates the seamless transfer of entity representations from the previous layer to the next one, promoting old knowledge preservation. Finally, we adopt a two-stage training paradigm to avoid the over-corruption of old knowledge influenced by under-trained new knowledge. Experimental results demonstrate the superiority of IncDE over state-of-the-art baselines. Notably, the incremental distillation mechanism contributes to improvements of 0.2%-6.5% in the mean reciprocal rank (MRR) score. More exploratory experiments validate the effectiveness of IncDE in proficiently learning new knowledge while preserving old knowledge across all time steps.

Supervised named entity recognition (NER) aims to classify entity mentions into a fixed number of pre-defined types. However, in real-world scenarios, unknown entity types are continually involved. Naive fine-tuning will result in catastrophic forgetting on old entity types. Existing continual methods usually depend on knowledge distillation to alleviate forgetting, which are less effective on long task sequences. Moreover, most of them are specific to the class-incremental scenario and cannot adapt to the online scenario, which is more common in practice. In this paper, we propose a unified framework called Contrastive Real-time Updating Prototype (CRUP) that can handle different scenarios for NER. Specifically, we train a Gaussian projection model by a regularized contrastive objective. After training on each batch, we store the mean vectors of representations belong to new entity types as their prototypes. Meanwhile, we update existing prototypes belong to old types only based on representations of the current batch. The final prototypes will be used for the nearest class mean classification. In this way, CRUP can handle different scenarios through its batch-wise learning. Moreover, CRUP can alleviate forgetting in continual scenarios only with current data instead of old data. To comprehensively evaluate CRUP, we construct extensive benchmarks based on various datasets. Experimental results show that CRUP significantly outperforms baselines in continual scenarios and is also competitive in the supervised scenario.

Large language models (LLMs) have demonstrated impressive proficiency in information retrieval, while they are prone to generating incorrect responses that conflict with reality, a phenomenon known as intrinsic hallucination. The critical challenge lies in the unclear and unreliable fact distribution within LLMs trained on vast amounts of data. The prevalent approach frames the factual detection task as a question-answering paradigm, where the LLMs are asked about factual knowledge and examined for correctness. However, existing studies primarily focused on deriving test cases only from several specific domains, such as movies and sports, limiting the comprehensive observation of missing knowledge and the analysis of unexpected hallucinations. To address this issue, we propose OntoFact, an adaptive framework for detecting unknown facts of LLMs, devoted to mining the ontology-level skeleton of the missing knowledge. Specifically, we argue that LLMs could expose the ontology-based similarity among missing facts and introduce five representative knowledge graphs (KGs) as benchmarks. We further devise a sophisticated ontology-driven reinforcement learning (ORL) mechanism to produce error-prone test cases with specific entities and relations automatically. The ORL mechanism rewards the KGs for navigating toward a feasible direction for unveiling factual errors. Moreover, empirical efforts demonstrate that dominant LLMs are biased towards answering Yes rather than No, regardless of whether this knowledge is included. To mitigate the overconfidence of LLMs, we leverage a hallucination-free detection (HFD) strategy to tackle unfair comparisons between baselines, thereby boosting the result robustness. Experimental results on 5 datasets, using 32 representative LLMs, reveal a general lack of fact in current LLMs. Notably, ChatGPT exhibits fact error rates of 51.6% on DBpedia and 64.7% on YAGO, respectively. Additionally, the ORL mechanism demonstrates promising error prediction scores, with F1 scores ranging from 70% to 90% across most LLMs. Compared to the exhaustive testing, ORL achieves an average recall of 80% while reducing evaluation time by 35.29% to 63.12%.

Named entity recognition (NER) aims to identify and classify specific entities mentioned in textual sentences. Most existing superior NER models employ the standard fully supervised paradigm, which requires a large amount of annotated data during training. In order to maintain performance with insufficient annotation resources (i.e., low resources), in-context learning (ICL) has drawn a lot of attention, due to its plug-and-play nature compared to other methods (e.g., meta-learning and prompt learning). In this manner, how to retrieve high-correlated demonstrations for target sentences serves as the key to emerging ICL ability. For the NER task, the correlation implies the consistency of both ontology (i.e., generalized entity type) and context (i.e., sentence semantic), which is ignored by previous NER demonstration retrieval techniques. To address this issue, we propose ConsistNER, a novel three-stage framework that incorporates ontological and contextual information for low-resource NER. Firstly, ConsistNER employs large language models (LLMs) to pre-recognize potential entities in a zero-shot manner. Secondly, ConsistNER retrieves the sentence-specific demonstrations for each target sentence based on the two following considerations: (1) Regarding ontological consistency, demonstrations are filtered into a candidate set based on ontology distribution. (2) Regarding contextual consistency, an entity-aware self-attention mechanism is introduced to focus more on the potential entities and semantic-correlated tokens. Finally, ConsistNER feeds the retrieved demonstrations for all target sentences into LLMs for prediction. We conduct experiments on four widely-adopted NER datasets, including both general and specific domains. Experimental results show that ConsistNER achieves a 6.01%-26.37% and 3.07%-21.18% improvement over the state-of-the-art baselines on Micro-F1 scores under 1- and 5-shot settings, respectively.

The in-context learning (ICL) for relational triple extraction (RTE) has achieved promising performance, but still encounters two key challenges: (1) how to design effective prompts and (2) how to select proper demonstrations. Existing methods, however, fail to address these challenges appropriately. On the one hand, they usually recast RTE task to text-to-text prompting formats, which is unnatural and results in a mismatch between the output format at the pre-training time and the inference time for large language models (LLMs). On the other hand, they only utilize surface natural language features and lack consideration of triple semantics in sample selection. These issues are blocking improved performance in ICL for RTE, thus we aim to tackle prompt designing and sample selection challenges simultaneously. To this end, we devise a tabular prompting for RTE (TableIE) which frames RTE task into a table generation task to incorporate explicit structured information into ICL, facilitating conversion of outputs to RTE structures. Then we propose instructive in-context learning (I^2CL) which only selects and annotates a few samples considering internal triple semantics in massive unlabeled samples. Specifically, we first adopt off-the-shelf LLMs to perform schema-agnostic pre-extraction of triples in unlabeled samples using TableIE. Then we propose a novel triple-level similarity metric considering triple semantics between these samples and train a sample retrieval model based on calculated similarities in pre-extracted unlabeled data. We also devise three different sample annotation strategies for various scenarios. Finally, the annotated samples are considered as few-shot demonstrations in ICL for RTE. Experimental results on two RTE benchmarks show that I^2CL with TableIE achieves state-of-the-art performance compared to other methods under various few-shot RTE settings.

Named Entity Recognition (NER) data augmentation (DA) aims to improve the performance and generalization capabilities of NER models by generating scalable training data. The key challenge lies in ensuring the generated samples maintain contextual diversity while preserving label consistency. However, existing dominant methods fail to simultaneously satisfy both criteria. Inspired by the extensive generative capabilities of large language models (LLMs), we propose ANGEL, a frAmework integrating the oNtoloGy structure and instructivE prompting within LLMs. Specifically, the hierarchical ontology structure guides prompt ranking, while instructive prompting enhances LLMs’ mastery of domain knowledge, empowering synthetic sample generation and annotation. Experiments show ANGEL surpasses state-of-the-art (SOTA) baselines, conferring absolute F1 increases of 2.86% and 0.93% on two benchmark datasets, respectively.

The immense parameter scale of large language models underscores the necessity for parameter-efficient fine-tuning methods. Methods based on Low-Rank Adaptation (LoRA) assume the low-rank characteristics of the incremental matrix and optimize the matrix obtained from low-rank decomposition. Although effective, these methods are constrained by a fixed and unalterable intrinsic rank, neglecting the variable importance of matrices. Consequently, methods for adaptive rank allocation are proposed, among which AdaLoRA demonstrates excellent fine-tuning performance. AdaLoRA conducts adaptation based on singular value decomposition (SVD), dynamically allocating intrinsic ranks according to importance. However, it still struggles to achieve a balance between fine-tuning effectiveness and efficiency, leading to limited rank allocation space. Additionally, the importance measurement focuses only on parameters with minimal impact on the loss, neglecting the dominant role of singular values in SVD-based matrices and the fluctuations during training. To address these issues, we propose SalientLoRA, which adaptively optimizes intrinsic ranks of LoRA via salience measurement. Firstly, during rank allocation, the salience measurement analyses the variation of singular value magnitudes across multiple time steps and establishes their inter-dependency relationships to assess the matrix importance. This measurement mitigates instability and randomness that may arise during importance assessment. Secondly, to achieve a balance between fine-tuning performance and efficiency, we propose an adaptive adjustment of time-series window, which adaptively controls the size of time-series for significance measurement and rank reduction during training, allowing for rapid rank allocation while maintaining training stability. This mechanism enables matrics to set a higher initial rank, thus expanding the allocation space for ranks. To evaluate the generality of our method across various tasks, we conduct experiments on natural language understanding (NLU), natural language generation (NLG), and large model instruction tuning tasks. Experimental results demonstrate the superiority of SalientLoRA, which outperforms state-of-the-art methods by 0.96\%-3.56\% on multiple datasets. Furthermore, as the rank allocation space expands, our method ensures fine-tuning efficiency, achieving a speed improvement of 94.5\% compared to AdaLoRA. The code is publicly available at https://github.com/Heyest/SalientLoRA.

Recent work for continual relation learning has achieved remarkable progress. However, most existing methods only focus on tackling catastrophic forgetting to improve performance in the existing setup, while continually learning relations in the real-world must overcome many other challenges. One is that the data possibly comes in an online streaming fashion with data distributions gradually changing and without distinct task boundaries. Another is that noisy labels are inevitable in real-world, as relation samples may be contaminated by label inconsistencies or labeled with distant supervision. In this work, therefore, we propose a novel continual relation learning framework that simultaneously addresses both online and noisy relation learning challenges. Our framework contains three key modules: (i) a sample separated online purifying module that divides the online data stream into clean and noisy samples, (ii) a self-supervised online learning module that circumvents inferior training signals caused by noisy data, and (iii) a semi-supervised offline finetuning module that ensures the participation of both clean and noisy samples. Experimental results on FewRel, TACRED and NYT-H with real-world noise demonstrate that our framework greatly outperforms the combinations of the state-of-the-art online continual learning and noisy label learning methods.

Knowledge distillation for knowledge graph embedding (KGE) aims to reduce the KGE model size to address the challenges of storage limitations and knowledge reasoning efficiency. However, current work still suffers from the performance drops when compressing a high-dimensional original KGE model to a low-dimensional distillation KGE model. Moreover, most work focuses on the reduction of inference time but ignores the time-consuming training process of distilling KGE models. In this paper, we propose IterDE, a novel knowledge distillation framework for KGEs. First, IterDE introduces an iterative distillation way and enables a KGE model to alternately be a student model and a teacher model during the iterative distillation process. Consequently, knowledge can be transferred in a smooth manner between high-dimensional teacher models and low-dimensional student models, while preserving good KGE performances. Furthermore, in order to optimize the training process, we consider that different optimization objects between hard label loss and soft label loss can affect the efficiency of training, and then we propose a soft-label weighting dynamic adjustment mechanism that can balance the inconsistency of optimization direction between hard and soft label loss by gradually increasing the weighting of soft label loss. Our experimental results demonstrate that IterDE achieves a new state-of-the-art distillation performance for KGEs compared to strong baselines on the link prediction task. Significantly, IterDE can reduce the training time by 50% on average. Finally, more exploratory experiments show that the soft-label weighting dynamic adjustment mechanism and more fine-grained iterations can improve distillation performance.

Low-resource relation extraction (LRE) aims to extract relations from limited labeled corpora. Existing work takes advantages of self-training or distant supervision to expand the limited labeled data in the data-driven approaches, while the selection bias of pseudo labels may cause the error accumulation in subsequent relation classification. To address this issue, this paper proposes fmLRE, an iterative feedback method based on feature mapping similarity calculation to improve the accuracy of pseudo labels. First, it calculates the similarities between pseudo-label and real-label data of the same category in a feature mapping space based on semantic features of labeled dataset after feature projection. Then, it fine-tunes initial model according to the iterative process of reinforcement learning. Finally, the similarity is used as a threshold for screening high-precision pseudo-labels and the basis for setting different rewards, which also acts as a penalty term for the loss function of relation classifier. Experimental results demonstrate that fmLRE achieves the state-of-the-art performance compared with strong baselines on two public datasets.

Recent work for extracting relations from texts has achieved excellent performance. However, existing studies mainly focus on simple relation extraction, these methods perform not well on overlapping triple problem because the tags of shared entities would conflict with each other. Especially, overlapping entities are common and indispensable in Chinese. To address this issue, this paper proposes PasCore, which utilizes a global pointer annotation strategy for overlapping relation extraction in Chinese. PasCore first obtains the sentence vector via general pre-training model encoder, and uses classifier to predicate relations. Subsequently, it uses global pointer annotation strategy for head entity annotation, which uses global tags to label the start and end positions of the entities. Finally, PasCore integrates the relation, head entity and its type to mark the tail entity. Furthermore, PasCore performs conditional layer normalization to fuse features, which connects all stages and greatly enriches the association between relations and entities. Experimental results on both Chinese and English real-world datasets demonstrate that PasCore outperforms strong baselines on relation extraction and, especially, shows superior performance on overlapping relation extraction.

Named entity recognition (NER) aims to locate and classify named entities in natural language texts. Most existing high-performance NER models employ a supervised paradigm, which requires a large quantity of high-quality annotated data during training. In order to help NER models perform well in few-shot scenarios, data augmentation approaches attempt to build extra data by means of random editing or by using end-to-end generation with PLMs. However, these methods focus on only the fluency of generated sentences, ignoring the syntactic correlation between the new and raw sentences. Such uncorrelation also brings low diversity and inconsistent labeling of synthetic samples. To fill this gap, we present SAINT (Syntactic-Aware InsertioN Transformer), a hard-constraint controlled text generation model that incorporates syntactic information. The proposed method operates by inserting new tokens between existing entities in a parallel manner. During insertion procedure, new tokens will be added taking both semantic and syntactic factors into account. Hence the resulting sentence can retain the syntactic correctness with respect to the raw data. Experimental results on two benchmark datasets, i.e., Ontonotes and Wikiann, demonstrate the comparable performance of SAINT over the state-of-the-art baselines.

A split based all-digital background calibration for an 18-bit pipelined-successive approximation register (SAR) analog-to-digital converter (ADC) is proposed in this paper. The "split ADC" architecture is exploited to eliminate the signal interference and accelerate the convergence speed of the calibration. The output difference of two channels continuously drives the background calibration algorithm to estimate and correct the error caused by nonlinearities. To guarantee the validity of calibration with different inputs, shuffling scheme is adopted in capacitor array. Different dither signals are injected into two channels to achieve faster convergence, especially under the DC input. In the meanwhile, enough redundancy is introduced to capacitor array to avoid the decrease of input range caused by dither signal. A 4-bit coarse ADC is exploited to alleviate conversion time and provide dither injection. A behavior level model with various nonideal factors is established in MATLAB to prove the availability of calibration for an 18-bit pipelined-SAR ADC. The simulation shows that the ENOB achieves 16.5-bit, the SNDR is increased from 52.4 dB to 100.9 dB and the SFDR is improved from 72.3 dB to 119.6 dB.

Recent work for extracting relations from texts has achieved excellent performance. However, most existing methods pay less attention to the efficiency, making it still challenging to quickly extract relations from massive or streaming text data in realistic scenarios. The main efficiency bottleneck is that these methods use a Transformer-based pre-trained language model for encoding, which heavily affects the training speed and inference speed. To address this issue, we propose a fast relation extraction model (FastRE) based on convolutional encoder and improved cascade binary tagging framework. Compared to previous work, FastRE employs several innovations to improve efficiency while also keeping promising performance. Concretely, FastRE adopts a novel convolutional encoder architecture combined with dilated convolution, gated unit and residual connection, which significantly reduces the computation cost of training and inference, while maintaining the satisfactory performance. Moreover, to improve the cascade binary tagging framework, FastRE first introduces a type-relation mapping mechanism to accelerate tagging efficiency and alleviate relation redundancy, and then utilizes a position-dependent adaptive thresholding strategy to obtain higher tagging accuracy and better model generalization. Experimental results demonstrate that FastRE is well balanced between efficiency and performance, and achieves 3-10$\times$ training speed, 7-15$\times$ inference speed faster, and 1/100 parameters compared to the state-of-the-art models, while the performance is still competitive. Our code is available at \url{https://github.com/seukgcode/FastRE}.

Previous entity linking methods in knowledge graphs (KGs) mostly link the textual mentions to corresponding entities. However, they have deficiencies in processing numerous multimodal data, when the text is too short to provide enough context. Consequently, we conceive the idea of introducing valuable information of other modalities, and propose a novel multimodal entity linking method with gated hierarchical multimodal fusion and contrastive training (GHMFC). Firstly, in order to discover the fine-grained inter-modal correlations, GHMFC extracts the hierarchical features of text and visual co-attention through the multi-modal co-attention mechanism: textual-guided visual attention and visual-guided textual attention. The former attention obtains weighted visual features under the guidance of textual information. In contrast, the latter attention produces weighted textual features under the guidance of visual information. Afterwards, gated fusion is used to evaluate the importance of hierarchical features of different modalities and integrate them into the final multimodal representations of mentions. Subsequently, contrastive training with two types of contrastive losses is designed to learn more generic multimodal features and reduce noise. Finally, the linking entities are selected by calculating the cosine similarity between representations of mentions and entities in KGs. To evaluate the proposed method, this paper releases two new open multimodal entity linking datasets: WikiMEL and Richpedia-MEL. Experimental results demonstrate that GHMFC can learn meaningful multimodal representation and significantly outperforms most of the baseline methods.

Sentiment and style control are two vital aspects in automatic poetry generation. Excellent Chinese classical poetry should express a certain emotion and embody a specific style at the same time. Existing work still has deficiencies in controlling sentiment and style simultaneously. To address above issues, in this paper, we propose a novel approach for Chinese classical poetry generation, which can generate sentiment-controllable and style-controllable poems. First, it classifies hundreds of thousands of poems by style, sentiment, format, and primary keyword. Then, it utilizes masking self-attention mechanism to associate multiple tags and verses. Besides, it can generate metrical rhyming verses with distinctive sentiment and style characteristics according to the tag-set and secondary keywords. Finally, this approach is applied in Chang Qing Yin, which can collaborate with users to polish generated poems, providing alternatives automatically. Experimental results show that our approach performs well in sentiment and style control, and quality of generated poems outperforms several strong baselines.

Knowledge graph embedding maps entities and relations into low-dimensional vector space.However, it is still challenging for many existing methods to model diverse relational patterns, especially symmetric and antisymmetric relations.To address this issue, we propose a novel model, AprilE, which employs triple-level self-attention and pseudo residual connection to model relational patterns.The triple-level self-attention treats head entity, relation, and tail entity as a sequence and captures the dependency within a triple.At the same time the pseudo residual connection retains primitive semantic features.Furthermore, to deal with symmetric and antisymmetric relations, two schemas of score function are designed via a position-adaptive mechanism.Experimental results on public datasets demonstrate that our model can produce expressive knowledge embedding and significantly outperforms most of the state-of-the-art works.

Recent years have seen an increasing emphasis on information security, and various encryption methods have been proposed. However, for symmetric encryption methods, the well-known encryption techniques still rely on the key space to guarantee security and suffer from frequent key updating. Aiming to solve those problems, this paper proposes a novel general symmetry-key encryption method based on deep learning called SEDL, where the secret key includes hyperparameters in deep learning model and the core step of encryption is processing input data with weights trained under hyperparameters. Firstly, both communication parties establish a weight vector table by training a deep learning model on the constructed synthetic training sets according to specified hyperparameters. Then, a self-update codebook is constructed on the weight vector table with the SHA-256 function and other tricks. When communication starts, encryption and decryption are equivalent to indexing the corresponding value on the codebook to obtain ciphertext or plaintext, respectively. Results of experiments and relevant analyses show that SEDL performs well for security, efficiency, generality, and has a lower demand for the frequency of key redistribution. Especially, as a supplement to current encryption methods, the time-consuming process of constructing a codebook increases the difficulty of brute-force attacks, meanwhile, it does not degrade the efficiency of communications.

This paper proposes a new vision-based method for extracting academic information from conference Web pages. The main contributions include: (1) An new vision-based page segmentation algorithm is proposed to improve the result of classical VIPS algorithm. This algorithm can divide pages into text blocks. (2) All text blocks are classified as 10 categories according to vision features, keyword features and text content features. The initial classification results have 75% precision and 67% recall. (3) The context information of text blocks are employed to repair and refine initial classification results, which are improved to 96% precision and 98% recall. Finally, academic information is extracted from classified text blocks. Our experimental results on real-world datasets show that the proposed method is effective and efficient for extracting academic information from conference Web pages.

Conference Web pages are the main platforms to share the conference information and organize conference events. To discover the academic knowledge from such Web pages for building academic ontologies or social networks, it is necessary to extract academic information from conference Web pages. This paper proposes an approach to extract academic information from conference Web pages. Firstly, Web pages are segmented into text blocks by analyzing the visual feature and DOM structure. Then Bayes Network is used to classify these text blocks into predefined categories, and the quality of initial classification results are improved after post-processing. Finally, the academic information is extracted from the classified text blocks. Our experimental results on the real world datasets show that the proposed method is highly effective and efficient for extracting academic information from conference Web pages, and it has average 90% precision and 89% recall.

Typical description logics are limited to dealing with crisp concepts and crisp roles. However, Web applications based on description logics should allow the treatment of the inherent imprecision. Therefore, it is necessary to add fuzzy features to description logics. A family of extended fuzzy description logics is proposed to enable representation and reasoning for complex fuzzy information. The extended fuzzy description logics introduce the cut sets of fuzzy concepts and fuzzy roles as atomic concepts and atomic roles, and inherit the concept and role constructors from description logics. The definitions of syntax, semantics, reasoning tasks, and reasoning properties are given for the extended fuzzy description logic. The extended fuzzy description logics adopt a special fuzzify-method with more expressive power than the previous fuzzy description logics.

Multiontologies environment is common such as the semantic annotation applications based on multiontologies. Based on analyzing and classifying the relationships between multiontologies, we propose the idea of bridge ontology, and present its detailed structure too. To deal with the semantic annotation problem based on multiontologies, a new approach is proposed which employing bridge ontology to express the complex relationships between multiontologies. The bridge ontology is a peculiar ontology, which can be created and maintained conveniently, and is effective in the applications based on multiontologies. The approach using bridge ontology has the advantages of low-cost, scalable, robust in the Web circumstance, avoiding the unnecessary ontology extending and integration, and promoting ontology reuse.

Multiontologies environments are common in semantic Web and knowledge representation applications. The key problem of these applications is how to represent the complex relationships between ontologies. For the purpose of solving the problem, This work analyzed and classified the relationships between multiontologies, then proposed the theory of bridge ontology. The bridge ontology is a peculiar ontology, has four-layer structures, and is able to express the twelve kinds of relations between multiontologies. Subsequently, on the premise of keeping the semantic consistency, the detailed discussions focused on the semantic checking and refining methods for the multiontologies with bridge ontology. Bridge ontology can be created and maintained conveniently, and is effective in the applications based on multiontologies. It has the advantages of low-cost, scalable, robust in the Web circumstance, avoiding the unnecessary ontology extending and integration, and promoting ontology reuse.