Retrieval-augmented generation (RAG) addresses the problem of knowledge cutoff and overcomes the inherent limitations of pre-trained language models by retrieving relevant information in real time. However, challenges related to efficiency and accuracy persist in current RAG strategies. A key issue is how to select appropriate methods for user queries of varying complexity dynamically. This study introduces a novel adaptive retrieval-augmented generation framework termed Layered Query Retrieval (LQR). The LQR framework focuses on query complexity classification, retrieval strategies, and relevance analysis, utilizing a custom-built training dataset to train smaller models that aid the large language model (LLM) in efficiently retrieving relevant information. A central technique in LQR is a semantic rule-based approach to distinguish between different levels of multi-hop queries. The process begins by parsing the user’s query for keywords, followed by a keyword-based document retrieval. Subsequently, we employ a natural language inference (NLI) model to assess whether the retrieved document is relevant to the query. We validated our approach on multiple single-hop and multi-hop datasets, demonstrating significant improvements in both accuracy and efficiency compared to existing single-step, multi-step, and adaptive methods. Our method exhibits high accuracy and efficiency, particularly on the HotpotQA dataset, where it outperforms the Adaptive-RAG method by improving accuracy by 9.4% and the F1 score by 16.14%. The proposed approach carefully balances retrieval efficiency with the accuracy of the LLM’s responses.

The accurate forecasting of vegetable prices is crucial for policy formulation, market decisions, and agricultural market stability. Traditional time-series models often require manual parameter tuning and struggle to effectively handle the complex non-linear characteristics of vegetable price data, limiting their predictive accuracy. This study conducts a comprehensive analysis of the performance of traditional methods, deep learning approaches, and cutting-edge large language models in vegetable-price forecasting using multiple predictive performance metrics. Experimental results demonstrate that large language models generally outperform other methods, but do not have consistent performance for all kinds of vegetables across different time scales. As a result, we propose a novel vegetable-price forecasting method based on mixture of expert models (VPF-MoE), which combines the strengths of large language models and deep learning methods. Different from the traditional single model prediction method, VPF-MoE can dynamically adapt to the characteristics of different vegetable types, dynamically select the best prediction method, and significantly improve the accuracy and robustness of the prediction. In addition, we optimized the application of large language models in vegetable-price forecasting, offering a new technological pathway for vegetable-price prediction.

Identifying frontier interdisciplinary domains is essential for tracking scientific evolution and informing strategic research planning. This study proposes a comprehensive framework that integrates (1) semantic disciplinary classification using a large language model (GPT-3.5-Turbo), (2) quantitative metrics for interdisciplinarity (degree and integration strength) and frontierness (novelty, growth, and impact), and (3) trend prediction using time series models, including Transformer, LSTM, GRU, Random Forest, and Linear Regression. The framework systematically captures both structural and temporal dimensions of emerging research fields. Compared to conventional citation-based or topic modeling approaches, it enhances semantic precision, supports multi-label classification, and enables forward-looking forecasts. Empirical validation shows that the Transformer model achieved the highest predictive performance, outperforming other deep learning and baseline models. As an illustrative example, the framework was applied to synthetic biology, which demonstrated high interdisciplinarity, strong novelty, and growing academic influence. These results underscore the field’s strategic position as a frontier interdisciplinary domain. Beyond this case, the proposed framework is generalizable to other domains and provides a scalable, data-driven solution for dynamic monitoring of emerging interdisciplinary areas. It holds promise for applications in science and technology intelligence, research evaluation, and policy support.

Feature engineering for whole-genome DNA sequences plays a critical role in predicting plant phenotypic traits. However, due to limitations in the models’ analytical capabilities and computational resources, the existing methods are predominantly confined to SNP-based approaches, which typically extract genetic variation sites for dimensionality reduction before feature extraction. These methods not only suffer from incomplete locus coverage and insufficient genetic information but also overlook the relationships between nucleotides, thereby restricting the accuracy of phenotypic trait prediction. Inspired by the parallels between gene sequences and natural language, the emergence of large language models (LLMs) offers novel approaches for addressing the challenge of constructing genome-wide feature representations with nucleotide granularity. This study proposes FE-WDNA, a whole-genome DNA sequence feature engineering method, using HyenaDNA to fine-tune it on whole-genome data from 1000 soybean samples. We thus provide deep insights into the contextual and long-range dependencies among nucleotide sites to derive comprehensive genome-wide feature vectors. We further evaluated the application of FE-WDNA in agronomic trait prediction, examining factors such as the context window length of the DNA input, feature vector dimensions, and trait prediction methods, achieving significant improvements compared to the existing SNP-based approaches. FE-WDNA provides a mode of high-quality DNA sequence feature engineering at nucleotide resolution, which can be transformed to other plants and directly applied to various computational breeding tasks.