Search Results (59)

Large Language Models (LLMs) have emerged as transformative tools across various domains, exhibiting remarkable capabilities in natural language processing and generation. However, their reliance on static pre-training data limits their ability to access up-to-date and domain-specific information. The existing research often treats augmentation strategies in isolation, and limited efforts have been made to systematically compare them through the lens of information integrity. This review focuses specifically on Retrieval-Augmented Generation (RAG) and fine-tuning, identifying them as the two dominant paradigms for integrating external knowledge: RAG for retrieval-based context injection and fine-tuning for parametric knowledge adaptation. While existing surveys predominantly focus on performance metrics like accuracy or latency, this paper addresses the critical gap of data fidelity—the preservation of truthfulness, integrity, and fairness during augmentation. We systematically synthesize empirical findings from diverse methodologies to determine how each approach mitigates hallucinations and bias. By comparing the trade-offs between retrieval-based context injection and parametric knowledge adaptation, this survey provides unique value to readers by providing a structured taxonomy, a unified evaluation framework, and actionable insights to guide future research and practical deployment of robust, high-fidelity LLMs. Full article

Metallurgical lifting equipment operates under prolonged heavy-load, high-impact, and complex working conditions. The resulting failure reports contain rich field knowledge applicable to fault diagnosis and predictive maintenance. Nevertheless, reliably extracting traceable, structured knowledge from procedural and implicit maintenance records remains a significant challenge. To address this, the paper proposes a generative information extraction method for large language models (LLMs) that integrates ontology schema with inference chain constraints, targeting knowledge extraction and knowledge graph construction from failure reports of metallurgical lifting equipment, named generative constrained information extraction for operations and maintenance (GCIE-OM). A domain ontology schema is first constructed, defining seven entity types and nine relation types to establish explicit knowledge boundaries for structured LLM generation. An inference chain-assisted structured parsing method, termed IC-ASP, is then designed to guide the model through a sequential extraction pipeline comprising scene identification, scope of entity boundary, inference of relation type, evidence traceability with localization, and triple output. This stepwise process strengthens the model’s capacity to comprehend equipment hierarchies, fault evolution chains, and maintenance action logic. Building on this, ChatGLM or LLaMA serves as the backbone model and is adapted to the target domain via LoRA fine-tuning. Entity alignment and character-level source localization mechanisms are further introduced to establish precise mappings between generated outputs and their textual evidence in the source documents. The extracted results are ultimately converted into standardized knowledge triples and stored in a Neo4j graph database. Based on this, a prototype system for generative information extraction is designed and implemented to demonstrate the practical effectiveness and adaptability of the proposed method. Experimental results show that the proposed method outperforms baseline methods across entity recognition, relation extraction, and structured output quality, providing robust knowledge support for fault tracing and predictive maintenance of metallurgical lifting equipment. Full article

Early wildfire monitoring in power transmission corridors requires reliable detection of weak fire and smoke cues under complex field conditions and strict edge-computing constraints. To address these issues, this paper proposes an edge-deployable RGB–thermal framework based on visible and thermal infrared (TIR) imaging for unmanned aerial vehicle (UAV)-based corridor monitoring, including a spatial detector, YOLO-MMSC, and a temporal-enhanced version, YOLO-MMSC-T. The study also establishes a self-collected corridor-oriented RGB–thermal (RGB–T) dataset to complement public wildfire data. Unlike existing RGB–thermal wildfire datasets that mainly focus on forest or wildland fire scenes, the proposed dataset is specifically organized for complex-background power transmission-corridor monitoring, including continuous UAV sequences, nighttime conditions, smoke/vegetation occlusion, long-range small targets, and hard-negative interference. To the best of our knowledge, this is the first self-collected RGB–thermal wildfire dataset designed for this specific application scenario. The framework integrates a mobile inverted bottleneck convolution (MBConv) lightweight backbone, a Shallow Detail Fusion Module (SDFM) for shallow cross-modal alignment and denoising, a Content-Guided Attention (CGA) module for adaptive fusion, and normalized Wasserstein distance (NWD)-based box regression for long-range small-target localization. Experiments on public and self-collected datasets show that YOLO-MMSC achieves 94.6% mAP@0.5, 95.0% precision, and 93.9% recall while running at 60 FPS on Jetson Orin NX. With temporal fine-tuning, YOLO-MMSC-T reaches a continuous detection rate (CDR) of 95.6% with a jitter index of $2.8\times {10}^{-3}$. Field experiments using a DJI Matrice 4T further indicate a practical operating altitude of 120–180 m. These results support lightweight RGB–thermal remote sensing for real-time wildfire monitoring in complex transmission-corridor environments. Full article

Background/Objectives: Disorganization of retinal inner layers (DRIL) is an important and supportive biomarker in optical coherence tomography (OCT) imaging for diagnosing the extent of diabetic macular edema (DME) in patients and anticipating visual outcomes. But the manual DRIL identification is subject to interobserver bias and requires a lot of time and effort from the experts. This research presents a novel, computerized, and clinically guided approach for the classification of DRIL that leverages the central 1 mm foveal region extracted through the annotations provided by the expert ophthalmologists and investigates the effectiveness of a transformer and Masked Auto Encoder (MAE) based foundation model (RETFound) as the primary approach. Methods: We fine-tuned and validated the RETFound model, utilizing accurate foveal center coordinates provided by the experienced ophthalmologists. Our approach emphasizes the macular region that is significant diagnostically, where DME biomarkers manifest more predominantly. To guarantee robust evaluation, the dataset was divided into 85% training and 15% held-out test sets. We performed 5-fold cross-validation exclusively on the training dataset with baseline, conservative, and moderate fine-tuning strategies, and the final model was evaluated on the independent, unseen test set. Convolutional neural network (CNN)-based transfer learning (TL) models (MobileNetV2, EfficientNetB0, InceptionV3, DenseNet121, and DenseNet169) were also assessed for comparative evaluation. Results: The RETFound model yielded the best outcomes under the conservative fine-tuning strategy, achieving a mean test accuracy (AC) of 0.9339 ± 0.0036 and an area under the curve (AUC) of 0.9660 ± 0.0028 on the independent held-out test set across the five fold-trained models. The moderate and baseline evaluations achieved comparatively lower outcomes, highlighting the effectiveness of the conservative approach. The RETFound model consistently outperformed CNN models, exhibiting stability and superior generalization for DRIL classification. We performed statistical validation using the Wilcoxon signed-rank test and 95% confidence intervals to confirm the robustness of the proposed method, and an ablation analysis showed that the fovea-centered region of interest (ROI) guidance consistently improved results when compared with whole OCT analysis. Conclusions: This research demonstrates that the deep-learning (DL) methods assisted by expert clinical knowledge with an anatomically aligned ROI could provide remarkable results in DRIL detection applications. This work attempts to establish an anatomically relevant framework for computerized DRIL identification that focuses on the highly crucial macular region, possibly helping in faster intervention and improved diagnosis in the management of DME. Full article

Accurate constitutive modeling of TC4 titanium alloy at elevated temperatures is critical for process design and numerical simulation in aerospace manufacturing. However, purely data-driven deep neural networks (DNNs) often suffer from severe overfitting and may yield physically unreasonable predictions in data-sparse or strictly out-of-distribution (OOD) regions. To address this issue, this study proposes a physics-guided two-stage neural network framework, termed NN-PhysicsInit, for the constitutive modeling of TC4 alloy. In Stage I, a large synthetic dataset generated from a strain-compensated Arrhenius-type constitutive equation is used to pre-train the network, thereby introducing analytical prior knowledge into the initial topological space. In Stage II, the pre-trained model is fine-tuned using rigorously corrected experimental data obtained from isothermal compression tests conducted over 800–980 °C and 0.001–1 s⁻¹ to improve material-specific predictive accuracy. To evaluate generalization capability, a rigorous dual-perspective extrapolation validation scheme is designed separately in the temperature (1010 °C) and strain-rate (10 s⁻¹) dimensions. The results demonstrate that, compared with direct black-box training, the proposed framework successfully prevents non-physical divergence and better preserves macroscopic thermodynamic smoothness in unseen domains. Specifically, the extrapolation average absolute relative error (AARE) is significantly reduced from 34.21% to 14.34% in the temperature extrapolation task, and from 27.91% to 8.92% in the strain-rate extrapolation task. These findings confirm that physics-based initialization acts as a powerful implicit regularizer, effectively mitigating the extrapolation catastrophe while maintaining high fitting accuracy. The proposed framework provides a robust and practical strategy for the constitutive modeling of complex alloys under limited-data conditions. Full article

The aircraft landing problem (ALP) poses significant challenges for traditional Monte Carlo Tree Search (MCTS) due to its vast search space and reliance on inefficient random simulations. To overcome these limitations, this paper proposes a novel Transformer-Augmented Monte Carlo Tree Search (TMCTS) algorithm. Our approach integrates a reinforcement learning framework that incorporates key operational constraints, including wake turbulence separation and time windows, and employs a cost function aimed at minimizing both delay time and fuel consumption. A core innovation is the replacement of the conventional random simulation phase in MCTS with a Transformer-based value predictor. This leverages the Transformer’s superior ability to model sequences and capture global dependencies among flights, thereby dramatically accelerating search convergence. Specifically, we designed a two-head Transformer network (comprising policy and value heads) to provide informed prior knowledge, which effectively guides the selection and expansion steps of the MCTS tree. The model is trained within an Actor–Critic framework, utilizing behavior cloning for pre-training followed by reinforcement learning for fine-tuning. Experimental evaluations on the standard OR-Library benchmark demonstrate that our TMCTS method significantly reduces scheduling deviation compared to state-of-the-art baselines (including FCFS, DPALO+GA, DPALO+PSO, and CPLEX). Moreover, it achieves a 93.7% reduction in computation time relative to the CPLEX method, highlighting its superior efficiency and practical applicability for real-time scheduling. Full article

Sentiment analysis is significant for exploiting countless opinion-rich data from social media. However, it faces known challenges, such as informal expression and data sparsity. While large language models (LLMs) excel at dynamic contextual disambiguation, such deep learning models neglect to leverage prior knowledge calculated statistically and globally based on dataset-wide category tendencies to guide the encoding process. To overcome these challenges, this study proposes a pre-attention framework that includes a modified gated recurrent unit hierarchically integrated with a dual-level pre-attention mechanism (PA-GRU) and an LLM-PA-GRU model equipped with a strategic LLM fine-tuning method. The pre-attention mechanism extracts inter-category and intra-text statistical priors from the training set. Trainable coefficient matrices are proposed to adaptively fuse these priors within a unified formulation, enabling flexible allocation of global and local statistical signals. The proposed PA-GRU cell features a modified gating structure and a computation flow that integrates these extracted priors as explicit statistical anchors during sequence encoding. Moreover, we implement an LLM-PA-GRU model that connects the LLM’s deep feature representations to the PA-GRU. This strategic LLM fine-tuning approach is intended to mitigate gradient instability frequently encountered in models with heterogeneous architectures. Finally, a prototype-based alignment loss is employed to enforce feature consistency across modules. Extensive experiments on benchmark datasets demonstrate that the proposed approach achieves competitive performance compared with recent literature-reported models. Full article

Due to the scarcity of high-quality, specialized datasets for coal mining equipment operation and maintenance (O&M) and the poor adaptability of large language models to domain-specific scenarios, the reliability of actual mining O&M cannot be guaranteed. To address this, this paper investigates the construction of vertical-domain large language models for coal mining equipment O&M scenarios under limited fine-tuning data. First, to tackle the lack of O&M scenario data, a safety-guided evolutionary self-instruction method (SafeEvol-Instruct), is developed by integrating Self-Instruction, Evol-Instruct, and Rule-Based Filtering. This approach achieves the unified fusion of scalable generation, deep evolution, and safety filtering on limited O&M data, resulting in the construction of scenario-specific datasets for system status assessment, equipment fault diagnosis, maintenance plan formulation, and preventive maintenance. Second, to account for the distinct characteristics of different O&M tasks, a hybrid fine-tuning strategy (SynergyLoRA) is proposed based on the Qwen2.5-7B-Instruct foundation model. This strategy incorporates middle-layer LoRA, top-layer LoRA, middle-layer IA³, Prompt Tuning, and Prefix Tuning to enable specialized training of vertical-domain models for each O&M scenario. Finally, the constructed Coal Mining Equipment O&M Large Language Model (CMEOM-LLM) is evaluated through ablation studies across various scenarios, validating the effectiveness of the proposed methods. Experimental results demonstrate that, in the system status assessment scenario, CMEOM-LLM achieves improvements of 4.9%, 1.5%, and 1.4% over the Qwen model in accuracy, recall, and F1-score, respectively. In the equipment fault diagnosis scenario, CMEOM-LLM outperforms Qwen by 7.4% in accuracy, with BLEU-4 and ROUGE-L scores increasing by 6.6% and 6.5%, respectively. In the maintenance plan formulation scenario, CMEOM-LLM surpasses ChatGLM with improvements of 6.6%, 6.5%, and 8.5% in ROUGE-L, BLEU-4, and human evaluation, respectively. In the preventive maintenance scenario, CMEOM-LLM achieves improvements of 7.1% and 8.9% over Qwen in ROUGE-L and BLEU-4, along with a 0.69-point increase in human evaluation scores. This paper provides an effective approach for knowledge management in coal mining equipment O&M. Full article

The digital transformation of oil and gas pipeline networks has generated substantial volumes of unstructured maintenance documentation from communication systems, creating an urgent need for automated summarization to improve operational efficiency. However, domain-specific text summarization for pipeline communication maintenance remains challenging due to scarce labeled data and the high computational cost of fine-tuning large pretrained models. Parameter-efficient fine-tuning alleviates this issue, but its effectiveness strongly depends on appropriate hyperparameter selection. This paper proposes a unified framework that combines weight-decomposed low-rank adaptation with an enhanced Artificial Bee Colony algorithm for automated hyperparameter optimization. The enhanced algorithm addresses two specific limitations of the standard Artificial Bee Colony algorithm: uninformed random initialization that ignores promising regions, and premature abandonment of stagnated solutions that discards partially useful search directions. These two components represent principled design choices, each targeting a distinct bottleneck in applying swarm intelligence search to high-dimensional mixed-type hyperparameter spaces. The method introduces a hybrid initialization strategy to exploit prior knowledge and a stagnation-guided local search mechanism to refine stagnated solutions instead of discarding them, achieving a better balance between exploration and exploitation. Experimental results on a public Chinese summarization benchmark and an industrial oil and gas pipeline communication maintenance corpus show that the proposed approach consistently outperforms full fine-tuning, manually tuned parameter-efficient methods, and several evolutionary optimization baselines in terms of ROUGE metrics. The automated search introduces modest additional computational overhead compared to manual tuning while eliminating expert-dependent hyperparameter configuration and achieving consistent performance gains across both datasets. Overall, the proposed framework provides an efficient and robust solution for adapting large language models to specialized summarization tasks in the context of pipeline communication system maintenance. Full article

Insect bites are a common cause of skin irritation and can contribute to disease transmission through vector-borne pathogens. Early identification of the likely biting organism can assist preliminary guidance (e.g., monitoring for warning signs, considering exposure history) and may reduce complications through timely follow-up. This paper studies a compact attention-guided learning framework for multiclass insect-bite image classification under strict storage constraints. A teacher network (BiteAI-T) based on MobileNetV3-Small is trained with spatial attention pooling to emphasize lesion-relevant regions while maintaining an efficient backbone. A lightweight depthwise-separable student (BiteAI-S) is trained using multi-level knowledge distillation that combines softened-logit matching with intermediate supervision through attention-map alignment and pooled-feature matching. Model storage is further reduced through weight-only quantization-aware training using an LSQ-inspired learnable scaling factor; BatchNorm running statistics are frozen during quantization fine-tuning to improve stability. Experiments on an eight-class dataset (ants, bed bugs, chiggers, fleas, mosquitos, no bites, spiders, ticks) show that BiteAI-T reaches 93.75% test accuracy. For deployment, we export (i) a TorchScript Lite teacher artifact (BiteAI-T_Lite, 2.35 MB) and (ii) a weight-only int8 student artifact (BiteAI-S_int8, 0.992 MB). Comparative results are also reported for an SVD-compressed + fine-tuned FP16 variant (92.66% test accuracy, 2.84 MB), illustrating accuracy–size trade-offs across compression strategies. Full article

Real-world Knowledge Graphs (KGs) are inherently incomplete, which hinders effective downstream reasoning. While Large Language Models (LLMs) possess powerful semantic capabilities, directly applying them to Knowledge Graph Completion (KGC) often leads to hallucinations and a lack of structural awareness. To address these challenges, we propose Embedding-Guided Instruction Tuning (EGIT), a novel framework that synergizes the structural precision of embedding models with the semantic reasoning of LLMs. Our approach operates in three key stages: (1) utilizing pre-trained embedding models to automatically synthesize high-quality, annotation-free instruction data; (2) fine-tuning the LLM with these structure-aware instructions to adapt it to the KGC task; and (3) employing a joint inference mechanism where the embedding model retrieves candidates and the fine-tuned LLM performs the final selection, thereby significantly reducing hallucinations. In extensive experiments, the best variant of EGIT achieves 7.0% and 2.5% improvements in Hits@1 on the FB15k-237 and WN18RR datasets, respectively. Full article

Destructive earthquakes frequently sever transportation lifelines, significantly impeding the progress of emergency rescue and post-disaster reconstruction efforts. The automated identification of road damage utilizing high-resolution remote sensing imagery is strictly constrained by the scarcity of post-disaster labeled samples and the morphological complexity of road networks. Consequently, model segmentation results frequently suffer from discontinuities in topological connectivity and confusion between background features and damaged roads. To address these challenges, this study proposes a road damage detection framework that integrates generative artificial intelligence with vector prior knowledge. A data simulation pipeline utilizing a stable diffusion model was constructed, employing topologically constrained masking to generate high-fidelity synthetic damage samples based on the DeepGlobe dataset, thereby mitigating the data deficit. The proposed Vector-Guided Damaged Road Segmentation Network (VRD-U2Net) employs wavelet convolutions (WTConv) to decouple high-frequency noise from low-frequency structural components and utilizes a Multi-Scale Residual Attention (MSRA) module to align visual features with vector priors. Furthermore, a vector-prior-driven dynamic upsampling mechanism is introduced to enforce geometric constraints on model predictions. Experimental results demonstrate that the method achieves an mIoU of 0.884 on the synthetic dataset. In validation using real-world imagery from the 2023 Turkey earthquake, the model attained an F1-score of 65.3% and recall of 72.3% without fine-tuning, exhibiting robust generalization capabilities to support manual damage assessment in data-scarce emergency scenarios. Full article

Ensuring that Large Language Models align with human values of honesty is a critical challenge, particularly due to the scarcity of labeled data for distinguishing known versus unknown knowledge boundaries. We propose a weak-to-strong generalization framework utilizing Group Relative Policy Optimization (GRPO). Unlike standard supervised fine-tuning or prompt engineering, our framework trains a lightweight “honest head” to rank response candidates based on multifaceted honesty scores. Crucially, we employ GRPO to optimize this head, leveraging group-relative advantages and PPO-style clipping to robustly learn from noisy, relative honesty signals. The weak honest head then guides the self-labeling of unlabeled data to fine-tune strong LLMs. Experiments on PopQA, SQuAD, Non-AmbigQA, and a domain-specific military medical dataset demonstrate that our framework significantly outperforms strong baselines, including Direct Preference Optimization (DPO), in honesty alignment. Full article

In nanometer integrated-circuit (IC) manufacturing, advanced technology scaling has intensified the effects of process variations on circuit reliability and performance. Random fluctuations in parameters such as threshold voltage, channel length, and oxide thickness further degrade design margins and increase the likelihood of functional failures. These variations often lead to rare circuit failure events, underscoring the importance of accurate yield estimation and robust design methodologies. Conventional Monte Carlo yield estimation is computationally infeasible as millions of simulations are required to capture failure events with extremely low probability. This paper presents a novel reliability-based circuit design optimization framework that leverages deep transfer learning to improve the efficiency of repeated yield analysis in optimization iterations. Based on pre-trained neural network models from prior design knowledge, we utilize model fine-tuning to accelerate importance sampling (IS) for yield estimation. To improve estimation accuracy, adversarial perturbations are introduced to calibrate uncertainty near the model decision boundary. Moreover, we propose a cascaded batch Bayesian optimization (CBBO) framework that incorporates a smart initialization strategy and a localized penalty mechanism, guiding the search process toward high-yield regions while satisfying nominal performance constraints. Experimental validation on SRAM circuits and amplifiers reveals that CBBO achieves a computational speedup of 2.02×–4.63× over state-of-the-art (SOTA) methods, without compromising accuracy and robustness. Full article

Knowledge graphs (KGs) offer a structured and collaborative approach to integrating diverse knowledge from various domains. However, constructing knowledge graphs typically requires significant manual effort and heavily relies on pretrained models, limiting their adaptability to specific sub-domains. This paper proposes an innovative, efficient, and locally deployable knowledge graph construction framework that leverages low-rank adaptation (LoRA) to fine-tune large language models (LLMs) in order to reduce noise. By integrating iterative optimization, consistency-guided filtering, and prompt-based extraction, the proposed method achieves a balance between precision and coverage, enabling the robust extraction of standardized subject–predicate–object triples from raw long texts. This makes it highly effective for knowledge graph construction and downstream reasoning tasks. We applied the parameter-efficient open-source model Qwen3-14B, and experimental results on the SciERC dataset show that, under strict matching (i.e., ensuring the exact matching of all components), our method achieved an F1 score of 0.358, outperforming the baseline model’s F1 score of 0.349. Under fuzzy matching (allowing some parts of the triples to be unmatched), the F1 score reached 0.447, outperforming the baseline model’s F1 score of 0.392, demonstrating the effectiveness of our approach. Ablation studies validate the robustness and generalization potential of our method, highlighting the contribution of each component to the overall performance. Full article