This study examines output divergence and club convergence among 34 newly established provinces in Vietnam from 2010 to 2023. Applying the Phillips and Sul (Econometrica 75:1771–1855, 2007; J Appl Econom 24:1153–1185, 2009)’s methodology to panel data at both aggregate and sectoral levels, the analysis reveals significant divergence in real GDP per capita across provinces. The results also identify multiple convergence clubs, with notable differences in club membership across economic sectors. At the aggregate level, four convergence clubs are observed. In a first-of-its-kind sectoral analysis for Vietnam, the study uncovers six convergence clubs in the Agriculture, Forestry, and Fishery sector, and four clubs each in the Industry and Construction, and Services sectors. Additionally, results from an ordered probit model show that initial GDP per capita, investment, labour force size, human capital, sectoral output shares, and population growth significantly influence club membership. Policy recommendations are derived from these findings.
To achieve agricultural automation, deep learning applications for early and accurate disease detection in tomato plants have been extensively developed. However, there is a fundamental trade-off between computational efficiency and diagnostic accuracy in resource-constrained agricultural edge environments. This paper proposes an evaluation framework for seven architectures that represent standard, efficient, and hybrid CNN structures to assess their implementation potential. Through evaluations of explainability, computational efficiency, and diagnostic performance, seven lightweight architectures (ShuffleNetV2, MobileNetV3-Small, SqueezeNet, MobilePlantViT, DenseNet121, ResNet50, and VGG16) are thoroughly examined. Three significant findings are derived from experiments conducted on a subset of tomato diseases in the PlantVillage dataset. First, the MobilePlantViT architecture accurately strikes the ideal balance between efficiency and performance. Second, in order to quantitatively assess the explainability of XAI models (Grad-CAM, SHAP, and LIME) and identify the best option for edge devices, we propose the perturbation stability score (PSS) metric. Third, we test CPU inference measurements to better reflect the actual scenario and find that the hybrid design effectively leverages parallel computing. According to these findings, MobilePlantViT is the ideal architecture for applications that require operation on edge devices with limited resources and achieve high diagnosis accuracy (above 99.5%).
Chilling injury (CI) is a primary limiting factor in the cold storage of eggplants (Solanum melongena L.), leading to rapid pericarp browning and quality deterioration. This study investigated the efficacy of folic acid (FA), a vital cofactor in carbon metabolism, as a novel postharvest strategy for alleviating CI and extending the marketable life and quality of eggplants during refrigeration. Freshly harvested eggplants were treated with FA at varying concentrations (0, 1.5, 3.0, 4.5, and 6.0 mg/L) and stored at 4 degrees C for 30 d. Every five days, the eggplants were subjected to various physicochemical and enzymatic analyses. The results demonstrated that FA application significantly delayed the onset of CI symptoms, effectively suppressing weight loss, electrolyte leakage, and malondialdehyde (MDA) accumulation in the tested samples. Enzymatic analysis revealed that FA treatment, particularly at 6.0 mg/L, modulated the enzymatic balance between oxidative defense and browning in the tested eggplants. Treatment maintained high activities of superoxide dismutase (SOD), catalase (CAT), and ascorbate peroxidase (APX), thereby reducing the oxidative stress burden under cold stress. Simultaneously, FA suppressed the activities of browning-related enzymes, polyphenol oxidase (PPO), and peroxidase (POD). Furthermore, FA upregulated phenylalanine ammonia-lyase (PAL) activity, leading to the preservation of total phenolics and flavonoids, which contributed to sustained non-enzymatic antioxidant capacities and radical-scavenging abilities (DPPH, ABTS, and FRAP). Principal component and Pearson correlation analyses confirmed that the FA-treated fruit maintained a biochemical profile distinct from that of the control samples. Among the tested FA treatments, eggplants treated with above 4.5 mg/L exhibited an excellent prevention effect from chilling injury. Overall, these findings suggest that FA mitigates CI by preserving membrane integrity and coordinating the antioxidant defense system, thereby offering a safe and effective approach to extend the shelf life of chilling-sensitive produce.
The complexity of Vietnam's legal texts presents a significant barrier to public access to justice. While Large Language Models offer a promising solution for legal text simplification, evaluating their true capabilities requires a multifaceted approach that goes beyond surface-level metrics. This paper introduces a comprehensive dual-aspect evaluation framework to address this need. First, we establish a performance benchmark for four state-of-the-art large language models (GPT-4o, Claude 3 Opus, Gemini 1.5 Pro, and Grok-1) across three key dimensions: Accuracy, Readability, and Consistency. Second, to understand the "why" behind these performance scores, we conduct a large-scale error analysis on a curated dataset of 60 complex Vietnamese legal articles, using a novel, expert-validated error typology. Our results reveal a crucial trade-off: models like Grok-1 excel in Readability and Consistency but compromise on fine-grained legal Accuracy, while models like Claude 3 Opus achieve high Accuracy scores that mask a significant number of subtle but critical reasoning errors. The error analysis pinpoints Incorrect Example and Misinterpretation as the most prevalent failures, confirming that the primary challenge for current LLMs is not summarization but controlled, accurate legal reasoning. By integrating a quantitative benchmark with a qualitative deep dive, our work provides a holistic and actionable assessment of LLMs for legal applications.
Reversible data hiding (RDH) has garnered significant attention from researchers due to its ability to completely restore the original images. absolute moment block truncation coding (AMBTC)-compressed images are widely used in RDH schemes. However, to obtain a large embedding capacity, several AMBTC-based RDH schemes use the decompressed pixels to carry the message bits. Therefore, stego-images cannot be decoded by traditional AMBTC decoders. In this paper, the new AMBTC-based RDH scheme is introduced to address this problem. The multi-pair value shifting technique is applied to embed data into the AMBTC code without changing the coding structure. Unlike existing AMBTC-based RDH methods that employ a large location map (LM) to ensure the correctness of the data extraction, our scheme can minimize the LM or avoid using it entirely by applying the multi-pair value shifting technique. The experimental results indicate that the proposed scheme achieves an appropriate embedding capacity while maintaining good visual quality.
