Yanchi Liu

The rapid rise of Large Language Model (LLM) agents is driving a fundamental paradigm shift in Multi-Agent Systems (MAS) research, moving from manually orchestrated static architectures toward automated configuration and optimization. Despite its significant potential, this frontier lacks a systematic and rigorous survey with clearly defined operational boundaries. To address this gap, this paper provides a comprehensive review of Automated MAS Optimization, formally anchoring it as the P4 paradigm within a six-stage evolutionary framework spanning from Foundation LLMs (P0) to Agentic Swarms (P5). We introduce precise mathematical definitions for core concepts, establishing a unified MAS configuration space that encompasses agent-level, system-level, and underlying components, and formulate the optimization objective as a holistic system-utility maximization problem. Furthermore, we partition P4 into three operationally distinct sub-paradigms based on the orthogonal dimensions of optimization timing and effect persistence: Design-Time Adaptive MAS, Test-Time Adaptive MAS, and Self-Evolving MAS. Guided by this taxonomy, we systematically review over 200 state-of-the-art works, covering both general methodologies and domain-specific applications. Beyond algorithmic perspectives, we critically examine key supporting issues including benchmarking, evaluation, and safety, while analyzing the evolutionary trajectory toward decentralized, emergent P5 Agentic Swarms. Finally, we identify core open challenges and propose future research directions centered on holistic configuration co-optimization, life-cycle evaluation, endogenous safety mechanisms, and the controllable transition from P4 to P5. This survey aims to provide a rigorous theoretical foundation and strategic navigation for researchers and practitioners in this rapidly evolving field.

Automatically extracting workflows as procedural graphs from natural language is promising yet underexplored, demanding both structural validity and logical alignment. While recent large language models (LLMs) show potential for procedural graph extraction, they often produce ill-formed structures or misinterpret logical flows. We present \model{}, a multi-agent framework that formulates procedural graph extraction as a multi-round reasoning process with dedicated structural and logical refinement. The framework iterates through three stages: (1) a graph extraction phase with the graph builder agent, (2) a structural feedback phase in which a simulation agent diagnoses and explains structural defects, and (3) a logical feedback phase in which a semantic agent aligns semantics between flow logic and linguistic cues in the source text. Important feedback is prioritized and expressed in natural language, which is injected into subsequent prompts, enabling interpretable and controllable refinement. This modular design allows agents to target distinct error types without supervision or parameter updates. Experiments demonstrate that \model{} achieves substantial improvements in both structural correctness and logical consistency over strong baselines.

Large language models (LLMs) have significantly advanced the field of natural language processing (NLP), providing a highly useful, task-agnostic foundation for a wide range of applications. However, directly applying LLMs to solve sophisticated problems in specific domains meets many hurdles, caused by the heterogeneity of domain data, the sophistication of domain knowledge, the uniqueness of domain objectives, and the diversity of the constraints (e.g., various social norms, cultural conformity, religious beliefs, and ethical standards in the domain applications). Domain specification techniques are key to making large language models disruptive in many applications. Specifically, to solve these hurdles, there has been a notable increase in research and practices conducted in recent years on the domain specialization of LLMs. This emerging field of study, with its substantial potential for impact, necessitates a comprehensive and systematic review to summarize better and guide ongoing work in this area. In this article, we present a comprehensive survey on domain specification techniques for large language models, an emerging direction critical for large language model applications. First, we propose a systematic taxonomy that categorizes the LLM domain-specialization techniques based on the accessibility to LLMs and summarizes the framework for all the subcategories as well as their relations and differences to each other. Second, we present an extensive taxonomy of critical application domains that can benefit dramatically from specialized LLMs, discussing their practical significance and open challenges. Last, we offer our insights into the current research status and future trends in this area.

The Internet of Things (IoT) is widely used in many applications such as smart city, transportation, healthcare, and environment monitoring. A key task of IoT maintenance is to analyze the abnormal sensor records and generate incident report. Traditionally, domain experts engage in such labor intensive tasks. Recent advances in Large Language Model (LLM) have sparked interests in developing AI based systems to automate these labor intensive processes. However, two critical problems hinder the effective application of LLM in IoTs: (1) LLM lacks background knowledge of deployed IoTs; and (2) the incidents are complex events involving many sensors and components. LLM needs to understand the sensor relationships for accurate diagnosis. In this study, we propose a Retrieval Augmented language model based Incident Diagnosing and Reporting system (RAIDR) for IoT applications. RAIDR retrieves related system documents based on the incident features and leverages LLM to analyze anomalies, identify root causes, and automatically generate incident reports. The automated incident reporting process streamlines end users’ decision making for system maintenance and troubleshooting.

Large Language Models (LLMs) exhibit potential artificial generic intelligence recently, however, their usage is costly with high response latency. Given mixed LLMs with their own strengths and weaknesses, LLM routing aims to identify the most suitable model for each query in the stream to maximize response quality and minimize cost and latency. However, the challenges involve: (1) dynamic trade-offs among quality, cost, and latency; (2) enabling continual learning in deployed systems; and (3) navigating a varying (e.g., new LLM addition or old LLM removal) set of LLM candidates over time. To bridge these gaps, we develop MixLLM, a dynamic contextual-bandit-based routing system for query-LLM assignment. Specifically, we first leverage query tags to enhance query embeddings for the routing task. Next, we design lightweight prediction models to estimate the response qualities and costs of queries over LLMs. We then devise a meta-decision maker to choose the query-LLM assignments to best tradeoff response quality, cost, and latency. Finally, the system benefits from continual training, allowing it to adapt to evolving queries and user feedback over time. Our extensive experiments show that MixLLM achieves the best trade-offs in response quality, cost, and latency (97.25% of GPT-4’s quality at 24.18% of the cost under the time constraint).

Imitation learning, which learns agent policy by mimicking expert demonstration, has shown promising results in many applications such as medical treatment regimes and self-driving vehicles. However, it remains a difficult task to interpret control policies learned by the agent. Difficulties mainly come from two aspects: 1) agents in imitation learning are usually implemented as deep neural networks, which are black-box models and lack interpretability; 2) the latent causal mechanism behind agents' decisions may vary along the trajectory, rather than staying static throughout time steps. To increase transparency and offer better interpretability of the neural agent, we propose to expose its captured knowledge in the form of a directed acyclic causal graph, with nodes being action and state variables and edges denoting the causal relations behind predictions. Furthermore, we design this causal discovery process to be state-dependent, enabling it to model the dynamics in latent causal graphs. Concretely, we conduct causal discovery from the perspective of Granger causality and propose a self-explainable imitation learning framework, CAIL. The proposed framework is composed of three parts: a dynamic causal discovery module, a causality encoding module, and a prediction module, and is trained in an end-to-end manner. After the model is learned, we can obtain causal relations among states and action variables behind its decisions, exposing policies learned by it. Experimental results on both synthetic and real-world datasets demonstrate the effectiveness of the proposed CAIL in learning the dynamic causal graphs for understanding the decision-making of imitation learning meanwhile maintaining high prediction accuracy.

Yijia Xiao, Yiqiao Jin, Yushi Bai, Yue Wu, Xianjun Yang, Xiao Luo, Wenchao Yu, Xujiang Zhao, Yanchi Liu, Quanquan Gu, Haifeng Chen, Wei Wang, Wei Cheng. Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing. 2024.

Fali Wang, Runxue Bao, Suhang Wang, Wenchao Yu, Yanchi Liu, Wei Cheng, Haifeng Chen. Findings of the Association for Computational Linguistics: EMNLP 2024. 2024.

Yufei Li, Xiao Yu, Yanghong Guo, Yanchi Liu, Haifeng Chen, Cong Liu. Findings of the Association for Computational Linguistics ACL 2024. 2024.

Chen Ling, Xujiang Zhao, Xuchao Zhang, Wei Cheng, Yanchi Liu, Yiyou Sun, Mika Oishi, Takao Osaki, Katsushi Matsuda, Jie Ji, Guangji Bai, Liang Zhao, Haifeng Chen. Proceedings of the 2024 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies (Volume 1: Long Papers). 2024.