Large Language Models (LLMs) have achieved strong performance on a wide range of complex reasoning tasks, yet further gains are often possible by leveraging the complementary strengths of multiple models. While multi-agent frameworks can improve solution quality by leveraging multiple LLMs, existing methods are often computationally expensive, both at training and inference time. In this work, we introduce a hierarchical multi-agent framework that addresses these challenges by training only a single leader LLM to coordinate a team of untrained peer agents. To this end, we propose Multi-agent guided Leader Policy \textbf{O}ptimization (MLPO), a novel approach which trains the leader to evaluate and synthesize agent responses without auxiliary value networks or explicit agent feedback. Leaders trained with MLPO exhibit improved performance not only when interacting with the agent team at inference time, but also enjoy improved performance when deployed in single-agent settings without the team. Empirical results on Big-Bench Hard (BBH), MATH, and MMLU demonstrate that our framework achieves substantial performance improvements over both single-agent and multi-agent baselines. Our results highlight the effectiveness and efficiency of training a single, flexible leader for collaborative reasoning in multi-agent LLM systems.

Modern vehicles, including autonomous vehicles and connected vehicles, have adopted an increasing variety of functionalities through connections and communications with other vehicles, smart devices, and infrastructures. However, the growing connectivity of the Internet of Vehicles (IoV) also increases the vulnerabilities to network attacks. To protect IoV systems against cyber threats, Intrusion Detection Systems (IDSs) that can identify malicious cyber-attacks have been developed using Machine Learning (ML) approaches. To accurately detect various types of attacks in IoV networks, we propose a novel ensemble IDS framework named Leader Class and Confidence Decision Ensemble (LCCDE). It is constructed by determining the best-performing ML model among three advanced ML algorithms (XGBoost, LightGBM, and CatBoost) for every class or type of attack. The class leader models with their prediction confidence values are then utilized to make accurate decisions regarding the detection of various types of cyber-attacks. Experiments on two public IoV security datasets (Car-Hacking and CICIDS2017 datasets) demonstrate the effectiveness of the proposed LCCDE for intrusion detection on both intra-vehicle and external networks.