As generalist LFMs approach diminishing returns from pure scaling, the industry is shifting toward multi-agentic systems, which are networks of small-sized specialized agents that coordinate tool use and decisions to solve complex tasks in vertical domains. Our research aims to construct multi-agent frameworks that not only achieve superior performance and efficiency in vertical scenarios but also prioritize safety and trustworthiness, laying the groundwork for reliable real-world deployment.
​
Multi-Agent Framework for Vertical Domain Adaptation: We aim to construct a parallel multi-agent system (MAS) that achieves long-term continuous self-evolve, fundamentally elevating the performance and efficiency of MAS beyond the limits of traditional LFMs. The core architecture encompasses a lead agent responsible for high-level planning and task orchestration, alongside specialized sub-agents tailored to domain-specific subtasks. A pivotal long-term goal is to build a high-quality multi-turn interactive multi-agent dataset that models long-horizon, multi-role collaborative trajectories, laying the foundation for self-evolve. Core long-term challenges include designing high-level planning and task decomposition mechanisms adaptive to complex vertical domain scenarios, developing evolutionary strategy learning methods, and exploring scalable post-training paradigms. Ultimately, we aim to realize a self-evolve multi-agent system capable of proactive optimization and adaptive improvement, enabling it to dynamically fit the evolving needs of vertical domains.
​
Collaborative Learning and Self-Reflection Capabilities for Agents: A core prerequisite for the long-term robustness and self-evolve of MAS lies in enhancing inherent collaborative learning and self-reflection capabilities of multi-agents. Specifically, we aim to enable multi-agents to effectively acquire knowledge and optimize strategies through interactive collaboration, while developing self-reflection mechanisms that allow them to learn from mistakes (e.g., decision-making errors, coordination failures). These capabilities are crucial for MAS to dynamically adapt to evolving vertical domain tasks and environments, continuously improve collaborative efficiency, and reduce repeated risks, thus laying a key foundation for the long-term reliable operation of multi-agent frameworks in practical vertical domain deployments.
​
MAS Evaluation for Trustworthy Vertical Deployment: A foundational long-term principle for building trustworthy MAS lies in acknowledging the inherent limitations of single-agent trustworthiness. We aim to propose a risk-graded governance framework: for high-risk decisions, human-in-the-loop or non-agentic models act as final arbiters to align behaviors with human intent, ethical norms, and industry safety standards; for low-to-medium risk scenarios, a committee of diverse agents suffices, optimizing resource allocation. Complementarily, we strive to develop domain-adaptive evaluation metrics to identify and mitigate hidden risks (e.g., domain-specific hallucinations, latent failures, and biases in information source selection) in MAS.