Recent advancements have underscored the exceptional analytical and situational understanding capabilities of Large Language Models (LLMs) in autonomous driving decisions. However, the inherent hallucination issues of LLMs pose significant safety concerns when utilized as standalone decision-making systems. To address these challenges, we propose the Hybrid-Driving framework, which leverages LLMs' situational comprehension and reasoning abilities alongside the specialized driving expertise embedded in knowledge graphs and driving rules, thereby enhancing the safety, robustness, and reliability of autonomous driving decisions. To articulate driving experiences clearly, we introduce the Scenario Evolution Knowledge Graph (SEKG), which integrates scenario prediction and action risk analysis in autonomous driving. By delineating observation areas and defining Time-to-Collision (TTC) levels, we effectively control the number of driving scenario nodes and ensure scenario diversity. Based on the scenario evolution relationships within the SEKG, we predict scenarios and assess associated action risks. Additionally, we implement a rule-filtering mechanism to eliminate unreasonable actions and employ prompt engineering to integrate scenario information, optional actions, and SEKG-based action risk analysis into the LLMs for decision-making. Extensive experiments demonstrate that our approach substantially improves decision success rates compared to using LLMs alone (≥37.5%), as well as surpasses the DiLu framework with LLMs and few-shot driving memory (≥7.5%), and other reinforcement learning methods (≥11%). These results validate the effectiveness of the Hybrid-Driving framework in enhancing LLM reliability for autonomous driving and advocate for its broader application of domain-specific knowledge across other fields.

The arterial blood pressure waveform (ABPW) serves as a less invasive technique for evaluating hemodynamic parameters, offering a lower risk compared to the more invasive pulmonary artery catheter (PAC) thermodilution method. Various studies suggest that deep learning models can potentially predict the hemodynamic parameters of ABPW. However, the scarcity of ground truth data restricts the accuracy of these models, preventing them from gaining clinical acceptance. To mitigate this data and domain challenge, this work proposed a self-supervised generative learning model for hemodynamic parameter prediction, called SSHemo (Self-Supervised Hemodynamic model). Specifically, SSHemo suggests first to leverage large amounts of unlabeled ABPW data to learn the representative embedding and then to fine-tune for the downstream task with a small amount of hemodynamic parameters’ ground truth. To verify the effectiveness of SSHemo, we utilize the public available VitalDB data set to train the model, and evaluation was conducted on two public datasets: VitalDB and MIMIC. The experimental results reveal that SSHemo’s regression mean absolute error (MAE) improved significantly from 1.63 L/min to 1.25 L/min when predicting cardiac output (CO). The trending tracking ability for CO changes meets clinical acceptance (radial limit of agreement (LOA) is $$\pm 25.56$$ °, less than $$\pm 30$$ °). In addition, SSHemo demonstrates robust stability in various conditions and cohorts, as evidenced by subgroup analysis, varying systemic vascular resistance (SVR) range analysis, and rapid CO analysis, compared to the most widely used commercial devices, the EV1000. Computational analysis further underscores the value and potential of practical application of the model in various settings.

The generation of adversarial safety-critical scenarios is essential for rigorously evaluating autonomous driving systems, enabling the identification of vulnerabilities and enhancement of system robustness. However, existing methodologies predominantly focus on extreme, unconstrained collision scenarios in which non-player character (NPC) vehicles exhibit unrealistic adversarial behaviors toward the ego vehicle. While such scenarios serve as stress tests, their practical utility is limited due to two key factors: 1) these extreme events are statistically rare in real-world traffic and frequently involve collisions that are physically unavoidable, irrespective of the autonomous vehicle’s decision-making capabilities; and 2) NPC behaviors in these scenarios are often intentionally aggressive (e.g., deliberate rear-end collisions), resulting in liability attribution that predominantly lies with the NPCs rather than exposing meaningful system limitations. Recent efforts to enhance scenario plausibility rely extensively on large-scale real-world traffic datasets, introducing significant computational costs and scalability constraints. To overcome these limitations, we propose a three-layer relative safety region model that partitions the driving environment into zones of varying risk levels. This partitioning increases the likelihood that NPC vehicles will interact within relative safety boundary regions, thus enabling the generation of more realistic and contextually relevant adversarial scenarios without the need for extensive real-world traffic data. We introduce AuthSim, a platform that integrates this safety model with reinforcement learning (RL) to generate both authentic and effective safety-critical scenarios. AuthSim is the first comprehensive approach to address both the authenticity and effectiveness of autonomous driving test scenarios without relying on large-scale traffic data. Empirical results demonstrate that AuthSim outperforms existing methods, achieving a 5.25% improvement in average cut-in distance and a 11.94% increase in average collision interval time compared with the state-of-the-art (SOTA) results, all while maintaining superior efficiency in scenario generation. These findings highlight the potential of AuthSim to produce high-fidelity and efficient test cases for the rigorous evaluation of autonomous driving systems.

This paper proposes a formal approach and testing process for the failure modes of intelligent algorithms. The process first requires the establishment of a failure mode library specific to the algorithm domain. Subsequently, we introduce a formalized method for providing a standardized and comprehensive description of the failure modes. We developed a supporting system to parse the formal description statements, enabling the automatic generation of failure test cases based on the failure modes. Finally, we validated the effectiveness and practicality of our proposed testing process based on failure modes in the field of object detection algorithms, demonstrating the testing process’s outstanding effectiveness.

With the rise of big data, the quality of datasets has become a crucial factor affecting the performance of machine learning models. High-quality datasets are essential for the realization of data value. This survey article summarizes the research direction of dataset quality in machine learning, including the definition of related concepts, analysis of quality issues and risks, and a review of dataset quality dimensions and metrics throughout the dataset lifecycle and a review of dataset quality metrics analyzed from a dataset lifecycle perspective and summarized in literatures. Furthermore, this article introduces a comprehensive quality evaluation process, which includes a framework for dataset quality evaluation with dimensions and metrics, computation methods for quality metrics, and assessment models. These studies provide valuable guidance for evaluating dataset quality in the field of machine learning, which can help improve the accuracy, efficiency, and generalization ability of machine learning models, and promote the development and application of artificial intelligence technology.