To address prevalent challenges in field-based wheat pest recognition—namely, viewpoint perturbations, sample scarcity, and heterogeneous data distributions—a pest identification framework named CropCLR-Wheat is proposed, which integrates self-supervised contrastive learning with an attention-enhanced mechanism. By incorporating a viewpoint-invariant feature encoder and a diffusion-based feature filtering module, the model significantly enhances pest damage localization and feature consistency, enabling high-accuracy recognition under limited-sample conditions. In 5-shot classification tasks, CropCLR-Wheat achieves a precision of 89.4%, a recall of 87.1%, and an accuracy of 88.2%; these metrics further improve to 92.3%, 90.5%, and 91.2%, respectively, under the 10-shot setting. In the semantic segmentation of wheat pest damage regions, the model attains a mean intersection over union (mIoU) of 82.7%, with precision and recall reaching 85.2% and 82.4%, respectively, markedly outperforming advanced models such as SegFormer and Mask R-CNN. In robustness evaluation under viewpoint disturbances, a prediction consistency rate of 88.7%, a confidence variation of only 7.8%, and a prediction consistency score (PCS) of 0.914 are recorded, indicating strong stability and adaptability. Deployment results further demonstrate the framework’s practical viability: on the Jetson Nano device, an inference latency of 84 ms, a frame rate of 11.9 FPS, and an accuracy of 88.2% are achieved. These results confirm the efficiency of the proposed approach in edge computing environments. By balancing generalization performance with deployability, the proposed method provides robust support for intelligent agricultural terminal systems and holds substantial potential for wide-scale application.

Short-term financial risk in intelligent trading systems is reflected not only in prices, trading volumes, and textual sentiment but also in infrastructure operating states, including server workload, device power consumption, network latency, and packet loss rate. We propose HSF-LLMNet, a hardware sensor and large language model semantic fusion network for jointly modeling external information shocks and infrastructure responses. A large language model extracts event category, sentiment polarity, risk intensity, and semantic uncertainty from financial texts. Reliability-aware temporal modeling handles sensor missingness, drift, and abnormal noise, while asynchronous soft alignment, bidirectional cross-attention, and reliability-aware gated fusion integrate irregular textual events with continuous hardware signals. The model jointly predicts market direction, realized volatility, and three-level risk over the subsequent 30 min. Experiments were conducted on eight Chinese A-share indices: the SSE Composite Index (000001.SH), SSE 50 Index (000016.SH), CSI 300 Index (000300.SH), STAR 50 Index (000688.SH), CSI 500 Index (000905.SH), CSI 1000 Index (000852.SH), Shenzhen Component Index (399001.SZ), and ChiNext Index (399006.SZ). The common observation period for market, textual, and hardware data extended from 1 March 2024 to 30 June 2025. After data cleaning, timestamp matching, and multimodal temporal alignment, 169,208 aligned asset–time prediction windows were retained for the 30 min forecasting task. Realized volatility was defined as the square root of the sum of squared one-minute log returns over the future 30 min interval. The three-level risk label was constructed from future realized volatility, absolute 30 min return, and liquidity stress, with all thresholds estimated exclusively from the training portion of each fold. A sample was labeled high risk when at least two of the three indicators exceeded their 85th-percentile thresholds or when any indicator exceeded its 95th-percentile threshold. It was labeled medium risk when, after excluding high-risk samples, at least two indicators exceeded their 60th-percentile thresholds or any indicator exceeded its 85th-percentile threshold; all remaining samples were labeled low risk. Results showed that HSF-LLMNet achieved an accuracy of 78.62%, a precision of 78.14%, a recall of 77.83%, an F1-score of 77.98%, an area under the receiver operating characteristic curve of 84.91%, and a Matthews correlation coefficient of 57.36% for directional prediction. For realized-volatility regression, the MAE, RMSE, MAPE, and $R^2$ were 0.0089, 0.0135, 9.21%, and 0.812, respectively. For high-risk-event warning, the mean effective warning time, defined as the interval between the first valid alarm and the corresponding event, was 15.37 min; the false-alarm rate and missed-alarm rate were 6.82% and 8.14%, respectively. Ablation experiments showed performance reductions after removing semantic encoding, sensor-reliability estimation, asynchronous alignment, bidirectional cross-attention, gated fusion, or multi-task learning. These results indicate that textual events and infrastructure operating states provide complementary information for quantitative risk analytics and fintech applications.

To address the challenge of collaborative strategy optimization caused by non-independent and identically distributed data in cross-institutional scenarios, a federated quantitative learning framework integrating Path Attention Aggregation Module (PAAM), Gradient Clipping and Compression (GCC), and a Heterogeneity-Aware Adaptive Optimizer (HAAO) is proposed to achieve efficient return optimization and robust risk control. The framework is validated across multi-market and multi-institutional environments, with experiments covering three key dimensions: return performance, risk management, and communication efficiency. The results demonstrate that the proposed model achieves an annualized return (AR) of 16.57%, representing an approximate 19.7% improvement over the traditional FedAvg model; the Sharpe ratio (SR) increases to 1.25, while the maximum drawdown (MDD) decreases to 15.92% and volatility remains controlled at 8.83%, indicating superior balance between return and risk. In the communication efficiency evaluation, when the number of communication rounds is reduced to 50 and 25, the model maintains accuracy at 94.2% and 92.8%, recall at 93.5% and 91.7%, and precision at 94.8% and 92.3%, respectively. Overall, the proposed framework achieves a dynamic balance between global convergence and risk constraints through path weighting, gradient sparsification, and frequency-domain learning rate adjustment. This research provides a novel and scalable paradigm for distributed financial prediction that ensures both privacy preservation and communication efficiency, demonstrating substantial engineering feasibility and practical applicability in intelligent financial modeling.