Battery electric buses (BEBs) are pivotal for sustainable urban transportation, yet their energy consumption is influenced by complex, interrelated factors that challenge accurate estimation and optimization. While machine learning models excel in energy prediction, their "black-box" nature limits practical deployment. This study addresses this gap by developing an interpretable machine learning framework integrating several machine learning models with SHapley Additive exPlanations (SHAP) and partial dependence plots (PDP). Using a high-fidelity simulation dataset of 169,344 scenarios generated via a validated MATrix LABoratory (MATLAB) Simulink model, we systematically analyze energy consumption under diverse driving conditions: covering extreme gradients (−8 % to 8 %), passenger loads (0–75), and Heating Ventilation and Air Condition (HVAC) usage (1.25–22.3 kW) as a proxy for temperature effects. The eXtreme Gradient Boosting (XGBoost) was selected as the best-performing machine learning model. SHAP analysis identified road gradient, initial state of charge (SoC), and Heating Ventilation and Air Condition (HVAC) usage as dominant factors, with nonlinear interactions between average speed and stop density ratio significantly impacting energy use. A human-machine interface (HMI) was developed to translate these insights into actionable recommendations for route optimization and driver training, enabling energy savings with minimal data acquisition costs. This study bridges the gap between theoretical energy models and practical decision-making, offering a robust framework for BEB fleet management while highlighting future directions for integrating real-world environmental data. • We fill the gap in the BEB energy consumption model interpretation. • We developed an interpretable Machine learning model for BEB energy consumption. • SHAP and PDP methods are combined to explain machine learning-based models. • Influence relationships and interactions of energy consumption factors are parsed. • Decision support tools and remarks are offered to fleet managers to reduce energy use.

The advancement of information technology and the rise of generative AI have paved the way for the development of Large Language Models (LLMs) tailored for TCM diagnostics. However, existing LLMs in the field of TCM face challenges in interpretability, limited modality in interaction, and robustness. To address these limitations, we propose MCM, a Multi-Agent Collaborative Multimodal Framework for TCM Diagnosis. This framework enables robust and interpretable multimodal diagnosis through multi-agent collaboration, offering novel methodologies for applying LLMs in the TCM domain. Experimental results demonstrate that the model within the MCM framework improved performance after fine-tuning, with additional capability gains under the MCM framework’s support, effectively addressing the challenges faced by LLMs in TCM, including interpretability, limited data modality, and lack of robustness. The code is open-sourced at: https://github.com/JerryMazeyu/MCM.

Large vision-language models (LVLMs) are vulnerable to harmful input compared to their language-only backbones. We investigated this vulnerability by exploring LVLMs internal dynamics, framing their inherent safety understanding in terms of three key capabilities. Specifically, we define these capabilities as safety perception, semantic understanding, and alignment for linguistic expression, and experimentally pinpointed their primary locations within the model architecture. The results indicate that safety perception often emerges before comprehensive semantic understanding, leading to the reduction in safety. Motivated by these findings, we propose Self-Aware Safety Augmentation (SASA), a technique that projects informative semantic representations from intermediate layers onto earlier safety-oriented layers. This approach leverages the model's inherent semantic understanding to enhance safety recognition without fine-tuning. Then, we employ linear probing to articulate the model's internal semantic comprehension to detect the risk before the generation process. Extensive experiments on various datasets and tasks demonstrate that SASA significantly improves the safety of LVLMs, with minimal impact on the utility.

Generative Adversarial Imitation Learning (GAIL) stands as a cornerstone approach in imitation learning. This paper investigates the gradient explosion in two types of GAIL: GAIL with deterministic policy (DE-GAIL) and GAIL with stochastic policy (ST-GAIL). We begin with the observation that the training can be highly unstable for DE-GAIL at the beginning of the training phase and end up divergence. Conversely, the ST-GAIL training trajectory remains consistent, reliably converging. To shed light on these disparities, we provide an explanation from a theoretical perspective. By establishing a probabilistic lower bound for GAIL, we demonstrate that gradient explosion is an inevitable outcome for DE-GAIL due to occasionally large expert-imitator policy disparity, whereas ST-GAIL does not have the issue with it. To substantiate our assertion, we illustrate how modifications in the reward function can mitigate the gradient explosion challenge. Finally, we propose CREDO, a simple yet effective strategy that clips the reward function during the training phase, allowing the GAIL to enjoy high data efficiency and stable trainability.

This paper focus on the polarimetric synthetic aperture radar (SAR) tomography for forested areas based on compressive MUSIC. In the proposed method, full polarimetric SAR echo signal reflected from the imaging area is collected, the corresponding multiple measurement vector model is established according to the parameters of polarimetric channels, the wavelet basis is adopted for representing the sparse vertical structure of the imaging area, and finally, the backscattering coefficients of the area are reconstructed by compressive MUSIC algorithm. The necessary number of tracks for SAR tomography is reduced and the severity of spurious spikes is suppressed under the same measurement accuracy. Simulation results from the PolSARpro data validate the effectiveness.