Ruifeng Guo

Large language models (LLMs) are advancing intelligent writing systems from local text continuation and language polishing toward long-form structured text generation. However, directly generating full-length academic paper drafts remains challenging due to unclear research objectives, unstable discourse structures, insufficient long-text coherence, and the lack of explicit quality control mechanisms. To address this long-form structured generation task, we propose MetricDraft, a metric-driven framework for academic paper draft generation. The framework organizes the drafting process as a closed-loop pipeline comprising research ideation clarification, structural anchoring, section-by-section generation, quality assessment, and feedback-driven revision. Its key components include adversarial research ideation clarification, staged structural anchoring, the PRISM structured metric system, progressive context injection with section-type-aware guided generation (PCI+STAGG), and a metric-feedback-driven generation–evaluation co-optimization mechanism. Experimental results demonstrate that MetricDraft achieves higher composite quality scores than one-shot generation, summary-based context passing, and context-accumulation-only baselines, improving MQS over Base1, Base2, and Base3 by +5.5, +7.9, and +7.0 points, respectively, with paired tests reaching statistical significance. To examine whether this advantage is tied to a single LLM backend, we further conduct a cross-model validation on all 15 tasks using Qwen3.7-Max in addition to the original DeepSeek-V4-Pro setting. MetricDraft remains the best-performing strategy under both models. To address citation reliability, an additional citation verification-and-retrieval-based replacement (CVRR) experiment reduces the fabricated citation rate of DeepSeek MetricDraft drafts from 56.0% to 15.0%. Furthermore, PRISM exhibits moderate-to-high positive correlations with expert ratings, providing preliminary evidence that it can serve as an auxiliary evaluation reference for draft quality diagnosis and iterative revision. This work reformulates academic writing as an adjustable, assessable, and iteratively optimizable long-form structured text generation problem, offering methodological insights for human–AI collaborative writing and intelligent text generation system design.

Large language models (LLMs) are advancing intelligent writing systems from local text continuation and language polishing toward long-form structured text generation. However, directly generating full-length academic paper drafts remains challenging due to unclear research objectives, unstable discourse structures, insufficient long-text coherence, and the lack of explicit quality control mechanisms. To address this long-form structured generation task, we propose MetricDraft, a metric-driven framework for academic paper draft generation. The framework organizes the drafting process as a closed-loop pipeline comprising research ideation clarification, structural anchoring, section-by-section generation, quality assessment, and feedback-driven revision. Its key components include adversarial research ideation clarification, staged structural anchoring, the PRISM structured metric system, progressive context injection with section-type-aware guided generation (PCI+STAGG), and a metric-feedback-driven generation–evaluation co-optimization mechanism. Experimental results demonstrate that MetricDraft achieves significantly higher composite quality scores compared to one-shot generation, summary-based context passing, and context-accumulation-only baselines, with differences reaching statistical significance. Furthermore, PRISM exhibits moderate-to-high positive correlations with expert ratings, providing preliminary evidence that it can serve as an auxiliary evaluation reference for draft quality diagnosis and iterative revision. This work reformulates academic writing as an adjustable, assessable, and iteratively optimizable long-form structured text generation problem, offering methodological insights for human–AI collaborative writing and intelligent text generation system design.

Bearing fault diagnosis in actual working conditions often faces the problem that unknown type faults cannot be identified, which seriously restricts the practical application of fault diagnosis technology. To solve this problem, this paper proposes a bearing fault diagnosis method based on transfer learning. Firstly, this paper designs a feature extraction network, the Multi-scale Convolution-Convolutional Reconstruction Network (MCRCNet), which incorporates a multi-scale feature extraction module to extract bearing fault features at multiple scales, thereby enhancing the extraction ability of key information. Secondly, this paper designs an improved convolutional reconstruction module AcConv (Adaptive Convolution reconstruction), which highlights key feature information and reduces redundant features by reconstructing the feature map. Furthermore, this paper also modifies the loss function to improve the performance in the case of data imbalance, and introduces the Wasserstein distance to optimize the adversarial training process. The proposed method is experimentally verified on Case Western Reserve University, Jiangnan University, and laboratory datasets. The experimental results show that the method has good performance in most tasks and has good generalization ability, which provides a feasible solution for the research of bearing fault diagnosis.

In advanced industrial automation, industrial robots have been widely utilized on assembly lines in order to reduce labor dependence. However, many related layout design approaches proposed are prone to generating unsafe layouts: there generally lacks a consideration regarding robots’ heights and assembly range, which will lead to costly collisions in the operation stage. In order to address the problem, we propose a three-dimensional (3D) optimization approach to a safe layout design for an assembly line with robots. We define modeling rules for robots to judge assembly ranges. A quantitative safety indicator is employed as a trigger for 3D collision detection in order to determine the positional relationship and status of the safe assembly collaboration. The optimization goals are established for minimizing the logistical cost and layout area in the model. A combined algorithm of differential evolution and nondominated sequencing genetic II is applied, which effectively enhances the poor diversity and convergence of the mainstream optimization method when solving this model. The benchmark tests and validation proved that our approach yields excellent convergence and distribution performance. The case study verifies that the safe layout model is valid and our approach can generate a safe layout in order to optimize economics and safety.