A comprehensive diagnosis of eddy–mean flow interaction in the Kuroshio Current (KC) region and the associated energy conversion pathway is conducted employing a state-of-the-art high-resolution global ocean–sea ice coupled model. The spatial distributions of the energy reservoirs and their conversions exhibit significant complexity. The cross-stream variation is found in the energy conversion pattern in the along-coast region, whereas a mixed positive–negative conversion pattern is observed in the off-coast region. Considering the area-integrated conversion rates between energy reservoirs, barotropic and baroclinic instabilities dominate the energy transferring from the mean flow to eddy field in the KC region. When the KC separates from the coast, it becomes highly unstable and the energy conversion rates intensify visibly; moreover, the local variations of the energy conversion are significantly influenced by the topography in the KC extension region. The mean available potential energy is the total energetic source to drive the barotropic and baroclinic energy pathway in the whole KC region, while the mean kinetic energy supplies the total energy in the extension region. For the whole KC region, the mean current transfers 84.9 GW of kinetic energy and 37.3 GW of available potential energy to the eddy field. The eddy kinetic energy is generated by mixed barotropic and baroclinic processes, amounting to 84.9 GW and 15.03 GW, respectively, indicating that topography dominates the generation of mesoscale eddy. Mean kinetic energy amounts to 11.08 GW of power from the mean available potential energy and subsequently supplies the barotropic pathway.

Despite the promise of graph neural networks (GNNs) in hydrological forecasting, existing approaches face critical limitations in capturing dynamic spatiotemporal correlations and integrating physical interpretability. To bridge this gap, we propose a spatial-temporal graph neural network (ST-GNN) that addresses these challenges through three key innovations: dynamic graph construction for adaptive spatial correlation learning, a physically-informed feature enhancement layer for soil moisture and evaporation integration, and a hybrid Graph-LSTM module for synergistic spatiotemporal dependency modeling. The temporal and spatial modules of the spatio-temporal graph neural network exhibit a structural symmetry, which enhances the model’s representational capability. By integrating these components, the model effectively represents rainfall-runoff processes. Experimental results across four Chinese watersheds demonstrate ST-GNN’s superior performance, particularly in semi-arid regions where prediction accuracy shows significant improvement. Compared to the best-performing baseline model (ST-GCN), our ST-GNN achieved an average reduction in root mean square error (RMSE) of 6.5% and an average improvement in the coefficient of determination (R²) of 1.8% across 1–8 h forecast lead times. Notably, in the semi-arid Pingyao watershed, the improvements reached 13.3% in RMSE reduction and 2.5% in R² enhancement. The model incorporates watershed physical characteristics through a feature fusion layer while employing an adaptive mechanism to capture spatiotemporal dependencies, enabling robust watershed-scale forecasting across diverse hydrological conditions.

Repository-level software engineering tasks are increasingly automated using repo-level retrieval-augmented generation (RLRAG), where a retriever selects relevant snippets from a repository to assist a language model (LM) in completing tasks. However, existing retrievers often lack effective designs to support LMs of varying capacities. To bridge this gap, we introduce RepoET, a novel retriever for RLRAG. RepoET organizes LMs and tools into an agentic workflow that closely mimics human search logic: “ search-files \(\to\) filter-files \(\to\) search-snippets \(\to\) filter-snippets ”. In our evaluation, RepoET outperformed state-of-the-art retrievers by accurately retrieving more relevant snippets in a better order across two widely used datasets, SWE-bench Lite and RepoQA, utilizing four LMs of different capabilities and sizes (as small as 3B). RepoET improved recall by over 12% and precision by over 20%, yielding a 21% improvement in Acc@k, along with superior snippet ordering. These retrieval improvements led to significant gains in downstream tasks: (1) we resolved 136 issues (45.33%) in SWE-bench Lite with GPT-4o without any additional information, and further improved the success rate by \(\sim\) 12% with multiple attempts; (2) we improved the accuracy of searching for designated functions by over 23% in RepoQA.