[Objective] In response to the prevalent issue of existing methodologies for measuring the similarity of research approaches in academic papers predominantly relying on unimodal features and their consequent limitations in identifying analogous research trajectories, this study proposes a novel approach for assessing the similarity of research methodologies based on multimodal feature alignment and fusion enhancement. [Methods] Initially, the image depicting the research framework is preprocessed using OpenCV for tasks including line detection and edge enhancement. Subsequently, a hierarchical vision transformer (Swin Transformer) and the Chinese-oriented optical character recognition model (CnOCR) are employed to extract visual and textual features from the diagram, respectively. Byte Pair Encoding tokenization and a lightweight LSTM model are utilized to derive contextual semantic embeddings. Finally, through tensor broadcasting and compact neural modules, the image and text features are aligned and non-linearly transformed. A learnable weighting mechanism integrates the multimodal representations, and retrieval and training are accomplished via dot-product similarity with batch-based classification loss.[Results] In the task of measuring the similarity of research idea graphs, the model proposed in this paper achieved 65.4%, 86.4%, and 88.2% in HR@1, HR@3, and HR@5 respectively, and simultaneously reached 82% and 83.3% in MAP and NCDG@5 metrics, outperforming various single-modal and multi-modal baseline models.[Limitations] The experimental dataset is exclusively confined to the domain of information resource management and does not encompass cross-language similarity detection. Moreover, the robustness of the model against complex image inputs remains to be enhanced.[Conclusions] The proposed multimodal feature alignment and fusion enhancement methodology significantly improves retrieval performance in research idea similarity measurement, providing a viable technical pathway for image-text multimodal-based paper similarity measurement, with potential applications in future tasks such as paper recommendation and plagiarism detection.

[Purpose] To address the issues of complex contract configuration and high negotiation costs for data circulation across multiple scenarios within trusted data space, this paper proposes a data circulation management method based on policy-based routing, aiming to enhance the intelligence of circulation decision-making while reducing costs and increasing efficiency. [Methods] Deployed at the contract pre-processing layer of the trusted data space service platform, this method constructs a five-dimensional tagging system by structurally characterizing the provider’s data attributes and the user’s requirements. It achieves the automatic matching of typical circulation scenarios based on a weighted scoring and rule arbitration mechanism. Ultimately, it generates a scenario-based contract configuration scheme, providing standardized input for digital contract generation and execution. [Results] Verification using the authorized operation of public data as an example demonstrates that, while remaining compatible with existing contracts and usage control constraints, the proposed method achieves the rapid matching and consistent configuration of data circulation schemes under multi-scenario conditions. This effectively improves the efficiency and operability of data circulation management. [Conclusion] This study converges complex circulation decisions into reusable scenario-based configuration rules and achieves intelligent matching in contract pre-processing stage. It enriches the functional architecture of the trusted data space and reduces transaction friction and compliance costs in scenarios such as the authorized operation of public data and government-enterprise collaboration.

[Objective] Developing a proactive identification and early warning system for academic integrity risks among researchers to provide theoretical support and methodological guidance for advancing research integrity.[Methods] Based on field theory, we constructed a dual-link proactive early warning framework for research integrity risks, designing and validating an integrated approach for risk identification, traceability evidence collection, and index calculation. First, we semantically linked and cleaned multi-source heterogeneous data on researchers according to CERIF standards to build a research collaboration knowledge graph. Second, we employed the DeepWalk algorithm for network representation learning and link prediction to uncover latent implicit benefit relationships. Third, integrating structural and textual features, a risk classification model was trained using the Random Forest algorithm to compute individual risk indices. Finally, empirical validation was conducted using retraction data from the National Natural Science Foundation of China's medical research domain, and a prototype system was developed to demonstrate the effectiveness of tiered and categorized early warning for researchers' integrity information.[Results] A risk identification model for researchers was established to generate an integrity risk index for researchers, and achieved potential risk assessment and warning of researchers proactively.[Limitations] The experimental data primarily focuses on the medical field, where academic integrity risks are relatively concentrated. The interdisciplinary generalizability of this method requires further validation. The tiered early warning index should be used as a supplementary reference and combined with expert experience.[Conclusions] Taking 200 researchers in the medical field as an example, this approach enables the quantitative identification and tiered early warning of academic integrity risks among researchers. A weighted calculation method is employed to determine their risk indices. The model achieved an AUC-ROC score of 0.716, precision-recall scores of 0.770, and an F1-score of 0.667. This demonstrates not only high interpretability and performance advantages but also features such as full traceability, evidence-based quantification, and tiered classification alerts. These capabilities validate the model's supportive role in assisting scientific research risk monitoring.

[Objective] To explore methods for identifying problem-solving collaboration patterns in online innovation communities, clarify the mechanisms through which different patterns influence innovation performance, and provide a theoretical foundation for optimizing platform management. [Methods] Based on a two-level argumentation framework, collaboration patterns were defined using graph models. Combining argument annotation with frequent set mining algorithms, an empirical analysis was conducted on 6,044 collaboration chains from the Salesforce community, identifying 12 micro-level and 24 macro-level collaboration patterns. On this basis, the impact of collaboration patterns on innovation performance and the mediating role of knowledge construction were tested. [Results] (1) Online innovation collaboration exhibits a two-tiered interactive process of “micro-level argument development—macro-level perspective integration”; (2) Stable collaboration patterns formed by sequences of argumentative activities exist in online innovation collaboration; at the micro-level, we identified non-developmental, low-evidence-development, and high-evidence-development patterns, while at the macro-level, we identified positive-development, mixed-development, and consensus-development patterns; (3) Different collaboration patterns exhibit distinct associations with innovation performance. At the micro level, the high-evidence development pattern exhibited the strongest effect（β=4.952，p<0.001）. At the macro level, the consensus development pattern（β=2.076，p<0.05） and the positive development pattern（β=0.936，p<0.001） significantly promoted innovation performance, while the mixed development pattern did not have a significant effect; Heterogeneity analysis indicates that the effects of different sub-structures vary; knowledge construction mediates the relationship between certain collaboration models and innovation performance. [Limitations] As this study is based on data from a single online innovation community, the temporal evolution of collaboration patterns and their cross-context applicability require further verification. [Conclusions] From a theoretical perspective, this study reveals the collaboration patterns in online innovation collaboration and their relationship with innovation performance, providing strategic insights for understanding collaboration processes within online innovation communities and optimizing collaboration on community platforms.

[Objective] Given problem statements, large language model is used to generate solution sentences to assist in discovering research ideas  and promoting scientific innovation.[Methods] A method based on group recommendation algorithm and supervised fine-tuning technique for large language models is presented to generate novel solutions for a problem. The keywords in the problem statement are extracted and inputed into the constrcuted recommendation model to obtain a set of solution keywords. Then, the problem statement and solution keywords are used to compose instructions that are inputed into the fine-tuned large language model to generate novel solutions.[Results] The proposed method can generate domain related and new solution sentences. Randomly select 100 new solutions and submit them to experts for evaluation. The average percentages of new solutions with novelty, referenceability, and comprehensibility are 70.25%, 74.5%, and 96.75%, respectively, indicating that this method can generate novel, valuable, and understandable solutions. [Limitations] The proposed method cannot output or utilize existing concepts that have not appeared in the paper to generate new solutions. Further validation is needed for experiments on multiple domain datasets and various types of large language models.

[Conclusions] The proposed method has to some extent solved the technical problem that the mined solutions are difficult to be understood in idea mining, and the generated new solutions have reference value, providing possibilities for accelerating scientific research discoveries or solving scientific research problems.

[Objective] Leverage a VAE-based Gaussian prior and vector quantization to fully capture the rich semantics and fine-grained nuances of words across diverse contexts.[Methods] We propose Gaussian Discrete Embedding (GDE): a variational autoencoder first expands static word vectors into a continuous Gaussian distribution to model semantic richness and hierarchy; vector quantization (VQ) then discretizes these continuous embeddings to reduce semantic entanglement and representation complexity; downstream tasks are finally handled by an RNN.[Results] GDE improves text-classification accuracy by 5–6 % and raises BLEU scores in machine translation by 2–3 points, confirming its effectiveness in capturing context-dependent word semantics.[Limitations]The experiments cover only a single language family,and the model's computational cost is high;extending to multilingual settings and improving efficiency are needed.[Conclusions]GDE achieves refined word semantic representation and significantly improves overall performance on downstream NLP tasks.

[Objective] To address the core issue of insufficient implicit relation recognition capability in document-level relation extraction, caused by information concentration or narrative perspective limitations in individual documents, and to enhance the model's relational reasoning performance in complex contexts lacking local co-occurrence clues. [Methods]This study proposes a document-level relation extraction method based on an Adaptive Category Memory Mechanism (ACME-RE). The method utilizes a memory module to dynamically maintain and update relation-specific prototype representations. Through a relation-guided contextual information quantification module, it independently assesses the token-level information contribution for each predefined relation, thereby enhancing the model's awareness and utilization of historical semantic prototypes and achieving cross-document semantic information enhancement.[Results]On the general dataset Re-DocRED, the proposed method achieves state-of-the-art (SOTA) performance with an F1 score of 84.61%. Compared to the strong general baseline ATLOP, it improves by 5.15%, maintains stable leading performance across all experimental settings, and outperforms other top methods by at least 1.3% in each test group. [Limitations] The method still exhibits certain limitations in processing extremely long documents, and its generalization capability requires further improvement. [Conclusions] This study validates the effectiveness of the adaptive category memory mechanism in enhancing the model's semantic reasoning ability and improving the performance of implicit relation extraction, providing new insights for document-level relation extraction research.

[Objective] This paper leverages embeddings from large pre-trained language models and multimodal medical knowledge enhancement techniques to improve the accuracy and semantic understanding capability of medical literature recommendation. [Methods] The study employs the LLM2Vec embedding model to achieve dynamic bidirectional encoding of medical texts, integrates a ResNet-50 visual encoder with an attention pooling mechanism for cross-modal feature fusion, and introduces a user behavior attention mechanism to capture interest evolution. A recommendation model is constructed through cross-modal alignment and multi-granular semantic aggregation. [Results] Comparative experiments with four mainstream recommendation models on two medical literature datasets show that the proposed model achieves an NDCG@10 of 0.579 on the PubMed/PMC dataset, which is 12.81 percentage points higher than the best baseline model LLM-ESR. On the Amazon books dataset, it attains an NDCG@10 of 0.548, outperforming LLM-ESR by 9.38 percentage points. These results demonstrate that the proposed model holds superior advantages in multimodal feature fusion and user interest modeling. [Limitations] The model's performance is highly dependent on the quality of multimodal data, and its computational complexity is relatively high. Performance may be compromised in data-sparse scenarios. [Conclusion] Experimental analysis confirms that the model effectively integrates multimodal medical knowledge and captures dynamic user interest evolution, significantly improving the accuracy of medical literature recommendation.

[Objective]To address role confusion and boundary misjudgment in prompt-based document-level event argument extraction models, this paper proposes a model with role-aware capsule modeling and prompt information gating. [Methods] The model adopts a RoBERTa-based encoder-decoder architecture, using structured prompts to generate argument candidates. Via a role-aware capsule network, it dynamically generates role-specific semantic representations with trigger words and candidate information. A prompt gating mechanism fuses structured prompts and context to enhance expressiveness; To improve matching accuracy, the model adopts the Hungarian matching algorithm and incorporates a span optimization strategy during training.[Results] Experiments were conducted on two mainstream event extraction datasets, RAMS and WikiEvents, F1-scores of 58.9% (Arg-I) and 53.8% (Arg-C) on RAMS, 71.8% (Arg-I) and 67.5% (Arg-C) on WikiEvents. [Limitations]The model still exhibits prediction biases in ultra-long texts or dense overlapping event scenarios. [Conclusions]Integrating role awareness and gating fusion in document-level event argument extraction effectively improves boundary prediction and role recognition accuracy, with experiments verifying its effectiveness and generality.

[Objective] To address guideline conflicts and hallucination risks of LLMs in multimorbidity management by proposing and validating a symbolic-layer-driven neuro-symbolic collaborative reasoning (SCGR) framework.

[Methods] A symbolic rule base using "IF-THEN-EXPLAIN" and a dynamic pruning algorithm were developed, followed by comparative experiments on 100 real clinical cases using phased Chain-of-Thought (CoT) for multi-dimensional evaluation.[Results] SCGR significantly outperformed LLM-only in semantic similarity (0.955 vs. 0.887) and clinical accuracy (5.0 vs. 4.08), effectively resolving decision deadlocks and demonstrating the value of logical transparency in establishing cognitive authority.[Limitations] The symbolic rule base has limited indicator coverage; complex multimorbidity interaction strategies require refinement; and the continuous expansion of the knowledge base remains partially dependent on manual maintenance.[Conclusions] SCGR aligns probabilistic models with clinical logic, providing a new paradigm for secure and traceable medical AI decision-making.

[Objective] This study explores the automatic generation and reasoning modeling methods for Traditional Chinese Medicine (TCM) medical case commentary, and verifies the effectiveness of the teacher-student collaborative reasoning and distillation framework in improving generation quality and clinical interpretability.[Methods] A dataset containing 510 complete medical cases was constructed based on authoritative TCM gynecology texts. Using DeepSeek-V3 as the teacher model and Qwen3-7B and Qwen3-14B as the student models, generative model distillation was implemented through reasoning chain samples generated by the teacher model. Parameter-efficient fine-tuning was performed using LoRA. The model performance was evaluated using BLEU, ROUGE, BERTScore, and paragraph-level ROUGE.[Results] Compared to the baseline model, model distillation improved the BLEU-4 score of Qwen3-14B from 16.62 to 32.77, and the ROUGE-L score from 0.148 to 0.39. With the introduction of data augmentation, the BLEU-4 score further increased to 35.11, and the paragraph-level ROUGE-L improved to 0.47. The model's alignment ability with the teacher's reasoning chain structure was significantly enhanced.[Limitations] The data primarily comes from gynecology case collections, with limited coverage of diseases and sample size. It has not yet been prospectively validated in real multi-center clinical settings.[Conclusions] The proposed teacher-student collaborative reasoning and distillation framework significantly improves the generation quality and clinical interpretability in small-sample TCM commentary generation tasks, demonstrating good feasibility and potential for wider application.

[Objective] This study aims to investigate the impact of academic inventors' knowledge integration capability on their patent output and to reveal the differential mechanisms of this impact across different types of academic inventors. [Methods] Based on publication and patent data in the biomedical field, a sample of 5,494 academic inventors was selected. A knowledge entropy model was constructed to measure the knowledge integration capability of different types of academic inventors, and its relationship with patent quantity and quality was analyzed. [Results] (1) The knowledge integration capability of research-oriented academic inventors was significantly higher than that of technology-oriented inventors, although the inter-group difference was limited; (2) Knowledge integration capability had a significant positive impact on patent quantity (β=0.083, p<0.05). Specifically, the knowledge integration capability of technology-oriented inventors significantly increased patent quantity (β=0.207, p<0.001); (3) Knowledge integration capability had a heterogeneous effect on patent citations: high integration capability of research-oriented inventors reduced patent citations, whereas that of technology-oriented inventors increased them.

[Limitations] The data of this study primarily originate from the biomedical field, and the generalizability of the conclusions needs to be verified in other disciplinary contexts. The categorization of academic inventors fails to fully capture the dynamic complexity of their careers. Furthermore, the measurement of knowledge integration capability is mainly based on explicit scientific publication data. [Conclusion] Knowledge integration capability generally contributes to higher patent quantity among academic inventors, but its effect on patent quality is significantly heterogeneous. This study reveals the differentiated relationships between knowledge integration and technological output across inventor types, and provides empirical evidence for classified talent evaluation and research commercialization management.

[Objective] This study aims to develop a multi-dimensional quantitative evaluation framework to precisely detect gender bias in Chinese large language models (LLMs) during recommendation-letter generation, and to further investigate its underlying causes and potential mitigation strategies. [Methods] Using corpora generated by DeepSeek-V3 and GLM-4, we employ odds ratios (OR) to assess lexical-content bias and build discrimination models based on BERT and RoBERTa. Gender differences in three language style indicators—sentiment intensity, proportion of formality, and confidence in formality—are evaluated via two-sided independent-samples t-tests and effect sizes (Cohen's d ). An NLI-based framework is used for hallucination detection, along with an assessment of hallucination amplification effects. [Results] In lexical content, LLMs exhibit inherent gender-representation imbalance in the CLG setting: “agentic” and “professional” words are significantly associated with males (maximum OR offset is +0.146), whereas the “private” words is significantly associated with females (maximum OR offset is -0.616). In language style, DeepSeek shows a female bias in sentiment intensity under CLG ( t =-3.21, p < 0.01), but after introducing background information (CBG), it shifts to a highly significant male bias ( t=6.51, p<0.001). The effect-size increase Δd in sentiment intensity reaches 0.436, indicating a pronounced male-bias amplification effect triggered by background information.[Limitations] This study focuses on only two mainstream Chinese models; the generalizability of the findings requires further validation. It also does not cover other potential bias dimensions (e.g., age, region).[Conclusions] The proposed multi-dimensional evaluation framework is shown to be effective and sensitive in measuring LLM bias. The findings suggest that bias in Chinese LLMs is highly covert and dynamic, and that high-quality background information alone is insufficient to eliminate deep-seated evaluative inequality. We propose a human-centered governance framework that enables life-cycle fairness control in LLM generation scenarios through counterfactual data augmentation, fairness-aware reward-function restructuring, and a multi-dimensional auditing system.

[Objective] This paper investigates the quantitative representation of relationship strength in power system knowledge graphs, aiming to address the limitations of traditional knowledge graphs that are confined to static binary relationships and fail to adequately capture the complex coupling characteristics and dynamic evolution patterns among power system entities.[Methods] To address the issues of insufficient relationship representation accuracy in power dispatch knowledge graphs, static modeling that overlooks system dynamic characteristics, and reasoning mechanisms lacking relationship strength awareness, this paper proposes a Coupling-aware Power Knowledge Graph (CPKG) construction method that embeds relationship coupling degrees. This method extends the traditional triple representation through quintuples <Entity1, Relationship Type, Entity2, Static Coupling Degree, Dynamic Coupling Degree>, designs multi-dimensional feature fusion algorithms to calculate static coupling degrees from electrical distance, topological connections, and functional correlations, and introduces dynamic adjustment mechanisms based on parameter correlation, state sensitivity, and historical association to achieve precise quantification and dynamic updating of relationship strength.[Results] Experiments conducted on real power dispatch system datasets validated the effectiveness of the CPKG method. In relationship characteristic analysis, the average coupling degrees of physical composition relationships and physical connection relationships reached 0.892 and 0.875, respectively. In reasoning task evaluation, CPKG achieved an F1 score of 94.05% in entity linking tasks, representing an 8.3% improvement over baseline methods, and an F1 score of 90.68% in relationship prediction tasks, showing a 7.2% improvement over baseline methods. Under complex fault scenarios, CPKG achieved a comprehensive accuracy rate of 84.6%, outperforming the best baseline method by 12.7 percentage points.

[Limitations] The CPKG method has currently been validated only on single regional power grid data, requiring extension to cross-regional interconnected power grid applicability studies.[Conclusions] By directly embedding relationship coupling degree information into knowledge graphs, the CPKG method effectively enhances the accuracy and dynamic adaptability of power system knowledge reasoning. This method provides crucial technical support for power system fault diagnosis, risk assessment, and operational optimization.

[Objective]Aiming at low-saliency events with implicit and ambiguously expressed temporal information in text, this paper proposes a relation extraction model named MGSSR, which is based on multi-granularity syntactic-semantic representation learning. [Methods]By analyzing contextual descriptions and integrating multi-source features to extract event arguments and tense information, this approach enhances event representations. Subsequently, temporal relation classification is performed based on the concatenated features of event pairs. [Results] Experiments were conducted on the MATRES dataset and TCR dataset, and the F1 values of the MGSSR model reached 82.8% and 85.9%, respectively, which are comparable to the performance of the optimal baseline model. [Limitations]The model's performance is somewhat dependent on upstream argument and tense extractors, and further improvement is needed in identifying relationships between cross-sentence events. [Conclusions] With enhanced event representations, our method enables more accurate temporal reasoning for low-saliency events.

[Objective]To dynamically fuse aspect information and global information and achieve cross-modal alignment at the distribution level, this paper proposes a dual-graph fusion enhancement and distribution alignment framework from a distributional perspective.[Methods]On the text side, a Dual-Graph Dynamic Fusion mechanism is employed to dynamically integrate aspect-related information with global contextual information, thereby strengthening text representations. In addition, a Gaussian–Contrastive Joint Optimization module is designed to realize cross-modal distribution alignment and enhance discriminative ability at the sample level.[Results]On the Twitter-2015 dataset, the DGDA model improves the highest accuracy and F1 score over the baseline models by 0.67% and 1.37%, respectively; on the Twitter-2017 dataset, the DGDA model achieves improvements of 0.87% in accuracy and 0.38% in F1 score compared with the baseline models.[Limitations]The model’s performance is, to some extent, constrained by the quality of affective knowledge and syntactic dependency construction. The utilization of visual information remains insufficient, and its generalization ability still needs to be validated on more datasets.[Conclusions]The proposed model dynamically fuses aspect information and global information and aligns text and images at the distribution level, while simultaneously enhancing the model’s discriminative capability.

[Objective] To capture the complex evolutionary relationships of disease progression and integrate incremental medical data, a link prediction-based risk assessment framework for new-onset multimorbidity in chronic diseases is proposed. [Methods] First, temporal triples were used to model disease onset pathways, and Multimorbidity Coefficient was introduced to quantify the risk of new-onset conditions and stratify risk levels. A Relational Graph Convolutional Network (R-GCN) was employed to learn representations of the disease and predict new-onset risks. Second, a cross-time window joint training strategy was designed. This strategy utilized embedded representations of historical data for warm-start model training, while incremental data were used to expand the encoder’s representation and integrated into the disease network, thereby iteratively optimizing the model through data integration. [Results] The R-GCN method achieved a maximum accuracy of 83.53% on the Hits@3 metric (i.e., the proportion of correct answers included in the top three predictions), outperforming representation learning and traditional neural network approaches. The joint training model RGCN-Jo++ shows a 2.42% improvement in Hits@1 on incremental data compared to historical data. [Limitations] The study is limited in data scope and volume, and rare diseases are not included. As the volume of new data increases, joint training consumes significant resources, and future work may incorporate incremental learning techniques. [Conclusions] The link prediction approach based on GNN provides robust support for predicting new-onset multimorbidity risk in chronic diseases.

[Objective] To construct an interpretable disease prognosis prediction model integrating knowledge graphs, in order to enhance the predictive ability and interpretability of disease prognosis prediction models. [Methods] The static features of patients are mapped to the entity nodes of the knowledge graph, and time series events are mapped to the relationship paths. The attention mechanism is adopted to integrate static, dynamic and graph structure embeddings, and the interpretability is evaluated in combination with SHAP analysis.[Results] Taking acute liver failure as the empirical field, the model achieved AUC 0.851, F1-score 0.663, and recall rate 0.684 on the test set; SHAP analysis showed that age, white blood cell count, lactate and liver and kidney function indicators were the key predictive features. [Limitations] The research data comes from a single center, and the generalization ability of the model in different disease types and multi-center clinical environments remains to be verified.[Conclusions] The model performs exceptionally well in short-term prognosis prediction and provides traceable structured explanations.

[Objective] This study aims to explore effective paths for enhancing the text generation capabilities of large language models in the field of financial research reports. [Methods] We propose an automatic summarization generation method for financial research reports based on a large language model. The main tasks include: 1) refining the key content dimensions of financial research reports using grounded theory; 2) constructing an instruction fine-tuning dataset based on the aforementioned dimensions; 3) adapting Qwen3-8B to the domain using LoRA technology, and evaluating the generated results from multiple dimensions. [Results] The fine-tuned model achieved F1 scores of 73.81%, 60.66%, and 69.75% on ROUGE-1, ROUGE-2, and ROUGE-L, respectively, surpassing the scores of 34.29%, 30.42%, and 31.82% for the unfine-tuned Qwen3-8B. The data results from both manual and intelligent evaluations further demonstrate that this method effectively enhances the quality of summaries. [Limitations] The current study only focuses on fine-tuning for a single text type of financial research reports, and the generalization ability of the model in other financial texts or cross-domain scenarios needs further verification and improvement. [Conclusions] The method proposed in this study can significantly enhance the generation quality of summaries for financial research reports, providing an effective reference path for the automated processing of structured texts.

[Objectives] Aiming at the problems of high difficulty in distinguishing deepfake claims and the poor interpretability of verification outcomes, this study explores the application of entailment tree and large language models in fact-checking tasks.. This will enhance capabilities for in-depth analysis of ambiguous content and transparent verification processes.[Methods] A LLM reasoner is used for chain reasoning to form natural language reasoning grounds and conclusions; a LLM transfer is employed to transform natural language into entailment trees for tracing ambiguity; and the LLM is fine-tuned to improve the quality of reasoning trees.

[Results] The proposed model can accurately identify deepfake claims, with precision rates of 60.60%、80.99%和89.99% on the AmbiFC, CHEF and FEVER datasets respectively, verifying its effectiveness in fact-checking tasks. [Limitations] The impact of the uncertainty of LLM output results on the overall model performance is not considered; multi-modal fact-checking issues are not taken into account.[Conclusions] The proposed model can effectively identify deepfake claims, provides efficient and implementable technical support for safeguarding cybersecurity in cyberspace.

[Objective] To address the issues of heterogeneous relational noise and insufficient global semantic modeling in the detection of highly human-like social bots. [Methods] To address these issues, this paper proposes a Transformer-based model with a relation graph retention mechanism, named Graph Retention-enhanced Transformer (GRT). It leverages a multi-scale retention mechanism to preserve high-quality structural information, introduces semantic neighborhood fusion to capture correlations between non-adjacent but semantically related nodes, and incorporates relation-level semantic attention to achieve effective cross-relation aggregation. [Results] On the TwiBot-20 and TwiBot-22 datasets, the accuracy and F1 scores of GRT are 0.8749, 0.8713 and 0.7622, 0.7621 respectively. Moreover, on TwiBot-20, compared with the suboptimal method, the accuracy is increased by 0.39% and the F1 score is increased by 1.66%. [Limitations] GRT still has room for optimization in scenarios with sparse structures and limited relational information.

[Conclusion] GRT effectively alleviates the problems of structural noise and insufficient semantic information in social networks, and realizes more accurate and robust detection of social robots.

[Objective] To address the challenges of scarce fine-grained image description data and high annotation costs, and to support the digitalization of intangible cultural heritage clay sculptures.[Methods] We proposes an image fine-grained description generation model, SKDP-VisualGLM, which parses images into structured triplets and, by incorporating key entity details and prompt engineering techniques, achieves large-scale, low-cost fine-grained description generation.

[Results] Experiments on the Tianjin “Niren Zhang” dataset show that, compared to baseline models such as Qwen2.5-VL-7B, the proposed model achieves stable improvements in metrics such as CN-CLIP (38.3 vs. 38.1), BLEU-2 (0.176 vs. 0.143), and BLEURT (0.636 vs. 0.625) compared to the best baseline.[Limitations] The multi-stage pipeline—from visual parsing to entity-level description and further to fine-grained description generation—still has room for improvement in terms of inference efficiency. In addition, the adaptability of the prompt engineering strategy to diverse image styles warrants further investigation.[Conclusions] By innovatively integrating global semantic structures with fine-grained key entity details, the SKDP-VisualGLM model effectively enhances the granularity of image descriptions, providing a feasible technical solution for the automated fine-grained description of intangible cultural heritage images such as clay sculptures.

[Objective]This study aims to address the contradiction between resource constraints and accurate recognition capabilities in large - model - based electronic medical record named entity recognition. It proposes a novel method called SlimNER to provide a new technical approach for intelligent medical record processing in primary healthcare settings.[Method]We selected the Qwen 3 - 0.6B model as the base and used the CCKS2019 dataset. The optimization integrated techniques like reasoning distillation, domain - adaptive fine - tuning, structurally enhanced attention mechanisms, and reinforcement learning. Specifically, we employed adaptive sparse attention mechanisms to capture sparse entities in long texts, applied chain - of - thought distillation and instruction distillation for knowledge transfer, and conducted staged fine - tuning to adapt the model to the medical domain and specific entity recognition tasks. All experiments were carried out in a resource - constrained notebook environment.[Results] On the CCKS2019 test set, the SlimNER method achieved an overall precision of 0.8439, a recall of 0.8745, and an F1 - score of 0.8588. Compared with recent studies, it showed high usability, especially with F1 - scores exceeding 0.85 for entities like diseases and diagnoses, and anatomical sites. However, it performed poorly in laboratory test - related entity recognition, with an F1 - score of only 0.7036.[Limitations]The SlimNER method has unsatisfactory performance in laboratory test - related entity recognition, mainly due to training data distribution bias, the vague boundaries of such entities, and the incompatibility of the adaptive sparse attention mechanism with these entities.[Conclusion]The SlimNER method can effectively improve the performance of lightweight large models in electronic medical record named entity recognition, making it suitable for resource - constrained environments and offering a new solution for primary healthcare settings. Yet, it still needs optimization for specific complex entity categories.

[Objective] Mobile applications (APPs) carry a large number of users’ private data, and privacy compliance analysis is a key factor in safeguarding user privacy. Given the inconsistencies among existing indicators for privacy policy compliance assessment, this study aims to construct a comprehensive compliance evaluation model based on multi-model fusion, thereby enhancing both the performance of privacy policy text classification and the comprehensiveness of compliance assessment indicators. [Methods] Based on multi-source privacy policy text data, this paper proposes a multi-model fusion privacy compliance analysis framework and implements a multi-metric comprehensive score for privacy policies. The text classification model uses the RoBERTa-TextCNN architecture, leveraging RoBERTa to obtain deep semantic embeddings and combining it with TextCNN to extract local key features, achieving a complementary advantage in privacy policy text classification. [Results] Experimental results show that the RoBERTa-TextCNN architecture achieves state-of-the-art classification performance on three public datasets. An empirical analysis was conducted on real-world privacy policy data from 2025, producing comprehensive compliance scores for different types of APPs and identifying key deficiencies in current compliance assessments. [Limitations] The weight of comprehensive score fails to consider the varying importance of different compliance indicators, and the evaluation indicators for AI-generated content applications require further refinement. [Conclusions] This paper successfully implements a comprehensive analysis framework for privacy compliance through multi-model fusion, thereby enhancing the coverage of privacy policy compliance indicators. By applying empirical analysis on real-world data, it provides government regulatory bodies with both data support and practical modeling tools, thereby promoting the formulation and optimization of privacy compliance management standards. The research holds significant theoretical value and practical implications.

[Objective] To address the problems of insufficient semantic information and inadequate cross-modal fusion in existing hate meme detection methods, this study proposes a detection method based on a semantic enhancement and dual cross-modal fusion architecture, aiming to improve detection accuracy and robustness.[Methods] The BLIP model is employed to generate image captions for enhancing textual semantics, while a dual cross-modal feature modeling mechanism is realized by combining a Dynamic Residual Fusion Module and an Exponential Weighted Fusion Module, and the fused features are finally classified using a cosine classifier to accomplish hate meme detection.[Results] On the PrideMM and Harm-P datasets, the proposed method achieves accuracy improvements of 1.0% and 2.8% over the second-best model, respectively, and significantly outperforms the baseline models. [Limitations] The model only utilizes image and text modalities, and thus does not cover hate content involving more complex modalities such as video and audio.[Conclusions] The proposed model leverages semantic enhancement and dual cross-modal fusion to effectively capture the deep semantic associations between image and text modalities, thereby improving the accuracy of hate meme detection.

[Objective] By introducing agents to align and enhance external knowledge, the representation of knowledge sparse conditions is optimized, and the problems of high similarity and inter-class ambiguity in the none-of-the-above relationships (NOTA) detection of few-shot relation classification are solved.[Methods] Our paper leverages concept graphs to address knowledge scarcity in low-sample scenarios and designs knowledge-augmented agents to dynamically align with external semantics. Furthermore, a mind tree with multiple reasoning paths is constructed to enhance the accuracy of knowledge injection, resulting in more precise prototype representations. This approach improves the accuracy of relation classification by enhancing the adaptability of NOTA detection within the prototype network.[Results] Our model proposed in this paper achieves performance improvements on both FewRel and Few-shot tacred: with the introduction of NOTA, it outperforms the current best methods by 2.55% and 2.92% respectively; without NOTA, the improvements are 0.34% and 1.36%. The ablation experiments also verify the effectiveness of each module of the model.[Limitations] The model in this paper needs to build a concept graph and query multi-hop concept sets based on the entity relationships in the dataset, which increases some construction costs and time overhead.[Conclusions] The model uses intelligent agents to optimize the traditional knowledge-enhanced prototype network, providing a new and highly accurate approach for few-sample relationship classification.

[Objective] To address the limited improvement in reasoning performance and the verbose output that arise when distilling complex chain-of-thought knowledge into small language models due to their capacity constraints, we propose a method named Segmented-Guided Knowledge Distillation (SGKD), which combines segmented-guided distillation with reasoning preference optimization. [Methods] SGKD first automatically segments long CoT into logically independent fragments, each augmented with a high-level semantic summary, to construct a structured knowledge representation. It then employs supervised fine-tuning to enable small models to efficiently absorb this structured knowledge and enhance their complex logical reasoning capabilities. Finally, it introduces a two-stage rejection sampling mechanism integrated with reasoning preference optimization to guide the generation of concise and correct reasoning paths. [Results] On mathematical reasoning benchmarks including GSM8K, MATH500, and AIME25, our method consistently outperforms existing baseline models. Specifically, compared with traditional long chain-of-thought distillation, SGKD achieves an average performance improvement of 2.27 percentage points on Qwen-series models and reduces the average output length of reasoning processes by 41.06 percentage points. [Limitations] The current approach does not examine its impact on non-reasoning abilities. [Conclusions] This work provides an effective and efficient solution for enhancing the complex reasoning capabilities of small language models, achieving notable performance gains while substantially improving output efficiency.

[Objective] To address the limitations of existing legal judgment prediction methods in handling complex interactions among defendants in multi-defendant cases, which leads to low prediction accuracy, a four-stage framework called MD-CRL Judge is proposed.[Method] First, a retrieval-augmented knowledge fusion mechanism is introduced. Second, a multi-agent collaborative analysis module is designed to analyze the case. Third, a multi-party debate-style reasoning approach is employed to deepen the analysis of facts and legal application. Finally, a judge agent synthesizes the case materials, retrieved knowledge, and debate records to output predicted charges for each defendant. [Results] Ablation and comparative experiments were conducted on the CMDL-small dataset. The experimental results demonstrate that the proposed model performs optimally, with accuracy, MP, MR, and F1 scores reaching 83.16%, 72.33%, 70.17%, and 69.48%, respectively. [Limitations] This method, based on multi-agent collaboration and debate, requires multiple calls to a large language model, resulting in relatively high inference overhead. [Conclusions] The MD-CRL Judge framework effectively simulates the judicial reasoning process, improves the accuracy of multi-defendant legal judgment prediction, and provides new approaches and technical support for smart justice.

[Objective] This study aims to compare the performance of different large language models and traditional rule-based or machine-learning methods in the task of fine-grained catalyst information extraction from full-text scientific literature, and to analyze the development trends in catalyst research based on the extracted results.[Methods] A manually annotated dataset comprising 194 full-text papers in the field of catalysis was constructed. Three large language models—DeepSeek, Gemini, and ChatGPT—along with the ChemDataExtractor toolkit were selected for performance comparison, and data analysis was conducted based on the extracted results.[Results] In fine-grained and complex catalyst information extraction, DeepSeek-v3.1-250821 achieved the best performance, with an overall average macro-F1 score of 0.88. Although ChemDataExtractor attained a high precision of 0.91, its F1-score was only 0.53, the lowest among the compared models. The data analysis further revealed several representative catalyst technology routes and key performance indicators.[Limitations] The experimental data in this study cover only English-language publications, and the scope of information extraction is limited to textual content. Some crucial information—such as catalyst stability metrics—was not effectively captured, as it is often presented in graphical form.[Conclusions] Comparative results across multiple models indicate that large language models combined with prompt engineering strategies exhibit overall advantages in catalyst information extraction, with DeepSeek-v3.1-250821 showing the best performance. These capabilities provide robust support for catalyst knowledge mining and technology assessment.

The study addressed the lack of joint analysis of asynchronous evolution in user behavior and social relationships with group sentiment polarization in cyberbullying detection. [Methods] A detection model integrating structural-difference features and group sentiment-polarization features was developed. Two types of discrete-time dynamic graphs were constructed to model user interaction and social-following relationships, respectively. A bidirectional graph convolutional network generated embeddings from both graphs, and structural-difference sequences were derived to characterize asynchronous structural evolution. A sentiment-confidence filtering mechanism produced positive and negative group sentiment-polarization sequences. An enhanced temporal-signal fusion model performed the final classification. [Results] Experiments on real-world social media datasets showed an accuracy of 89.5% and an F1-score of 87.0%, with improvements of about 2% to 7% over representative baseline models. [Limitations] This study was validated solely on specific social media datasets. Future work could incorporate data from more diverse platforms and test the model in real-world environments to assess its generalization capability and robustness. [Conclusions] The fusion of social network structure difference and group sentiment polarization model based on group polarization perspective can significantly improve the detection effect of cyberbullying.

[Objective] This study aims to integrate heterogeneous information across countries, markets, and modalities to overcome the limitations of single data sources and enhance the prediction accuracy of China's new energy stock indices.[Methods] We develop a Deeply Coupled Long Short-Term Memory Attention Model (DC-LSTM-Attention-LLM). The model employs ChatGPT with structured zero-shot prompting to quantify investor sentiment from text-based information; constructs an LSTM ensemble to learn deep representations from cross-country, cross-market, and multimodal data; incorporates a multi-head attention mechanism to capture interaction dependencies among heterogeneous features; and performs feature fusion via a shared layer with ReLU activation to generate index predictions.[Results] Empirical analyses on Chinese and U.S. new energy indices show that DC-LSTM-Attention-LLM consistently outperforms nine benchmark models across all evaluation metrics. Specifically, the model achieves an average 12.83% performance improvement relative to the standard LSTM, demonstrating its superiority in forecasting complex financial time series.[Limitations] The accuracy of sentiment recognition using zero-shot prompting is constrained when dealing with complex financial semantics. Future research will introduce advanced prompt engineering to enhance the model's recognition capability and robustness.[Conclusions] The deep modeling approach integrating cross-country, cross-market, and multi-modal data can effectively capture complex market characteristics and significantly enhance the forecasting accuracy of China's new energy stock market.

[Objective] To address the difficulty of achieving prospective identification and early intervention of clinical misdiagnosis, this study proposes an approach integrating prior knowledge with multi-source heterogeneous data for misdiagnosis risk prediction and feature identification, aiming to improve the timeliness and operability of misdiagnosis detection. [Methods] Prior misdiagnosis knowledge was extracted from expert experience and clinical rules to assist in determining misdiagnosis events and to enable knowledge-driven label construction; a hybrid machine learning model was built based on electronic medical records, with interpretable learning introduced to identify key risk features. Model development used data from 19,256 patients provided by the China National Population Health Data Center, and external evaluation and a clinical pilot were conducted on an independent validation cohort of 2,153 patients from Peking University People’s Hospital. [Results] In independent validation, the model achieved an accuracy of 92%, an AUROC of 0.90, and an AUPRC of 0.67; the clinical pilot results showed that the Youden index increased from 84.50% to 92.65% after model deployment. [Limitations] The construction of misdiagnosis labels relies on rules and expert knowledge, and the pilot sample size and duration are limited; generalizability and robustness still require multicenter and longer follow-up validation. [Conclusions] The proposed approach achieves favorable performance in misdiagnosis risk prediction and key feature identification, enhances early intervention capability and decision-support value in real clinical practice, and provides a feasible technical pathway for building an intelligent prevention and control system for clinical misdiagnosis.

[Objective]To improve the accuracy and timeliness of identifying technological innovation partners for enterprises, this paper proposes a link prediction method based on a Dynamic Graph Convolutional Network (DGCN) with multi-modal feature fusion.[Methods]This paper proposes an attention-based model for fusing the topological, domain, and semantic features of nodes in a patent collaboration network. A GCN-LSTM architecture is designed with a sliding time window strategy to capture the network's dynamic evolution. The model performs link prediction to identify potential technological innovation partners.[Results]An empirical study on a patent dataset from China's new energy vehicle (NEV) sector from 2015 to 2024 shows that our method significantly outperforms baseline models across all metrics. It achieved an AUC of 0.858, an improvement of 5.0 percentage points over the next-best model, EvolveGCN, and an F1 score of 0.807, which is 3.5 percentage points higher than the runner-up model, DySAT.[Limitations]The study did not fully exploit patent-specific features such as citation relationships and patent value. Furthermore, it did not integrate non-patent, multi-source heterogeneous information, such as corporate R&D investment and market performance.[Conclusions] By effectively capturing the dynamic evolutionary patterns of patent cooperation networks and comprehensively utilizing multimodal patent features, this method can provide more precise and forward-looking data-driven decision support for enterprises engaging in collaborative innovation.

[Objective] To break down disciplinary barriers, explore common knowledge between disciplines, and build a bridge for cross-field dialogue.[Methods] Analyze the citation context of existing interdisciplinary common knowledge cases to extract co-citation patterns, then extract these patterns and combine indicators of commonality, interdisciplinarity, and novelty to identify common knowledge with potential value.[Results] 1,044 pairs of candidate interdisciplinary knowledge pairs were identified from 53 million citation records, and an in-depth analysis was conducted on 7 cases with high scores in interdisciplinarity, commonality, and novelty, revealing the intersections of disciplinary issues at the basic methodological level.[Limitations] The number of syntactic patterns currently used needs to be expanded, and there is still room for improvement in the acquisition of interdisciplinary common knowledge instances.[Conclusions] The constructed method based on citation context analysis provides a new path paradigm and methodology for the common connection between disciplines.

[Objective] This study aims to address the challenges of data utilization efficiency and privacy protection in multi-institutional data fusion, thereby enhancing both the practicality and security of data sharing among multiple institutions.[Method] A novel multi-institutional data fusion model, termed CEFDFM-MI (Cloud-Edge Federated Data Fusion Model for Multi-Institution), is proposed, which integrates cloud-edge collaboration with federated learning. Within the cloud-edge collaborative framework, a budget allocation mechanism and an information gain evaluation mechanism are devised to ensure the efficiency of data fusion processes. Furthermore, by leveraging the distributed characteristics of federated learning, the proposed model ensures secure and privacy-preserving data integration across institutions. The model is empirically evaluated on MNIST, CIFAR-10, and CIFAR-100 datasets under independent and identically distributed (IID) scenarios, as well as non-IID scenarios exhibiting low, medium, and high degrees of heterogeneity, in order to assess its performance across diverse and complex environments.[Results] Under IID conditions, the CEFDFM-MI model achieves a maximum accuracy of 94.52%. In non-IID scenarios with low, medium, and high heterogeneity, the model attains peak F1 scores of 73.71%, 74.51%, and 73.45%, respectively. Moreover, in the presence of model heterogeneity at the edge level, the proposed model demonstrates an accuracy improvement of approximately 6%–8% compared to independent training on individual edge nodes.[Limitations] The current study does not address scenarios where the objectives of cloud and edge models are misaligned, and the model's applicability in more complex environments remains to be further explored.[Conclusions] The proposed CEFDFM-MI model exhibits superior global performance relative to FedAvg and FedProx, and possesses robust capabilities in handling model heterogeneity across multiple institutions.

[Objective] In view of the characteristics of centralized storage, data security risks, limited computing resources, and urgent user analysis and utilization needs of the National Population Health Data Center (NPHDC), this study explores the construction ideas suitable for the data enclave of NPHDC, so as to provide users with a more efficient, secure, and flexible data processing and analysis environment.[Methods] The types, characteristics, implementation mechanisms, and applicability of different scenarios of data enclaves were summarized. Combined with the data application characteristics of NPHDC, a big data analysis platform for NPHDC was built based on the virtual enclave method integrating of security enhancement, micro-isolation, and artificial intelligence technologies.[Results] The big data analysis platform supported services such as data review, data processing, data analysis and mining, and peer review of the data associated with the user's published papers in NPHDC. It has completed the review tasks of 32,000 datasets of more than 2,800 projects, more than 10,000 data analysis tasks, and more than 5000 data processing tasks, with a data leakage rate of 0% and a resource utilization rate of 80%.[Limitations] It is not possible to realise cross-institutional data sharing with decentralized storage, and it is necessary to explore data enclave research combining privacy-preserving technologies such as multi-party secure computing and federated learning in combination with the development of NPHDC.[Conclusions] It is of great significance to effectively solve the needs of safe sharing and collaborative analysis of population health data centralisation, which is of great significance for the security and sharing and utilisation of national population health scientific data.

[Objective] To enhance enterprises' situational awareness and emergency decision-making abilities through a situational awareness model for sudden incidents.[Methods] This paper establish a model from perception and understanding dimensions. It utilizes knowledge graphs and deep learning for event knowledge extraction, constructs a corporate emergency symbiotic system based on symbiosis theory and Logistic growth model. The " Earth Pit Pickled Mustard Greens" incident is used as a case study.[Results] Construct the model for the " Earth Pit Pickled Mustard Greens " incident, followed by an analysis of the situational development stages and the relationships among situational elements,  providing situational assessment recommendations. [Limitations] Data quality and knowledge graph structure need improvement.[Conclusions] The proposed model helps extract situational elements and identify relationships, enhancing emergency decision-making capabilities.

[Objective]Performing named entity recognition of core technical information in patent texts with few labeled samples.[Methods]Utilizing the features of large language model, which possess abundant general knowledge and significant semantic understanding capabilities, we propose a methodological framework for named entity recognition from patent text using prompt template.[Results]This paper presents an empirical analysis using the TFH-2020 annotated dataset pertinent to hard disk drive. The experimental results demonstrate that with the few-shot learning ability of large language model, the performance of named entities achieves 68% in F1-value, while with the supervised fine-tuning method,it can achieve to 54%, which is contrary to its performance on general texts.[Limitations]Although the proposed method significantly reduces the cost of data annotation, there is still a considerable gap compared to deep learning methods using a large amount of labeled data for model training. In addition, the design and optimization methods of Prompt templates, as well as the technology for rapidly generating large batches of instruction sets, need to be further improved.[Conclusions] Compared to random sample selection strategy, the NER performance of large language model using similar sample selection strategy has increased from 29% to 69% (in F1 value). This indicates that sample selection strategy has a significant impact on the performance of large language models in patent NER task, and prompt template is the core of this method, it not only determines the quality of the recognition effect but also influences the choice of optimization methods.

[Objective]  Entity relationship extraction models in general domains have poor performance when directly applied to specific domains. Statistical analysis reveals that entity location, type and entity relationship in Tujia ethnic instrumental music texts have strong correlation characteristics, and the paper proposes an entity relationship extraction model that integrates entity location and type characteristics.[Methods] Adopting the Pipeline relationship extraction model, after completing the named entity recognition task, the relative position of each character to the subject-object and entity type features are spliced into the original relationship statement, and then the features are learned through the BERT model, and finally the relationship classification is learned through the fully connected layer. [Results]  The ablation and model comparison experiments on the self-constructed Tujia ethnic instrumental music dataset show that the model (BERT_E) that incorporates the entity type features performs optimally, with an F1-micro of 97.359%.[Limitations]  The sample size is small, and the entity location features do not take entity length into account.[Conclusions] The research results promote the digital protection and intelligent application services of Tujia ethnic instrumental music culture, and also have important reference value for entity relationship extraction in ethnic instrumental music related fields.

[Objective]Traditional recommendation methods can not effectively generate personalized recommendations when facing the cold start problem, which reduces the accuracy of recommendation and user satisfaction. [Methods]Different from the general task of recommending financial commodities only using commodity information, this paper also carries on the characterization of user history transaction records and introduces multi-level fusion characterization. By linking the two, a recommendation system that incorporates implicit information can capture complex user investment patterns. [Results]Firstly, different benchmark methods were compared, and then a small sample learning scenario was constructed to verify the ability of this model to deal with the commodity cold start problem. The experimental results show that compared with the previous optimal method, the average reciprocal ranking, hit rate and normalized discount cumulative gain of this method are increased by 18.6%, 26.08% and 23.52%, respectively. [Limitations]Since it is not based on the most advanced neural network architecture, more advanced deep neural networks can be used in the future to further improve the recommendation effect.[Conclusions] The results of different ablation experiments and comparison with the benchmark model prove the effectiveness of the proposed method in recommending financial commodities.