Large Language Model-based Multi-Agent Systems (LLM-based MAS), where multiple LLM agents collaborate to solve complex tasks, have shown impressive performance in many areas. However, MAS are typically distributed across different devices or environments, making them vulnerable to perturbations such as agent failures. While existing works have studied the adversarial attacks and corresponding defense strategies, they mainly focus on reactively detecting and mitigating attacks after they occur rather than proactively designing inherently resilient systems. In this work, we study the resilience of LLM-based MAS under perturbations and find that both the communication topology and prompt design significantly influence system resilience. Motivated by these findings, we propose ResMAS: a two-stage framework for enhancing MAS resilience. First, we train a reward model to predict the MAS’s resilience, based on which we train a topology generator to automatically design resilient topology for specific tasks through reinforcement learning. Second, we introduce a topology-aware prompt optimization method that refines each agent’s prompt based on its connections and interactions with other agents. Extensive experiments across a range of tasks show that our approach substantially improves MAS resilience under various constraints. Moreover, our framework demonstrates strong generalization ability to new tasks and models, highlighting its potential for building resilient MASs.

With the rapid advancement of medical artificial intelligence, Traditional Chinese Medicine (TCM) has demonstrated significant potential in intelligent prescription recommendation. However, the scarcity of structured and high-quality clinical data often leads to issues such as data sparsity and terminology inconsistency, which limit the generalization capability and practical performance of existing models. To tackle these challenges, this study proposes PresRecSKG, a knowledge graph-guided data augmentation framework that utilizes a herbsymptom knowledge graph to enhance both multi-label and multi-class prescription recommendation tasks. The core of the framework is the proposed Sampling-by-Knowledge-Graph (SabKG) strategy, which effectively incorporates external knowledge for data synthesis, thereby mitigating data sparsity and improving data quality control. Experimental results show that PresRecSKG achieves the best performance in the multi-label task, with an average improvement of approximately 2.5% across all evaluation metrics. In the multi-class task, the framework elevates performance from 0.94 to over 0.97 across key metrics. These findings indicate that PresRecSKG offers an effective data augmentation paradigm for TCM prescription recommendation, enhancing the accuracy, compatibility, and interpretability of intelligent TCM systems.

Recent advancements in Large Language Models (LLMs) have led to a rapid growth of agentic systems capable of handling a wide range of complex tasks. However, current research largely relies on manual, task-specific design, limiting their adaptability to novel tasks. In this paper, we introduce a new research problem: Modularized LLM Agent Search (MoLAS). We propose a modular design space that abstracts existing LLM agent designs into four fundamental modules with uniform IO interface: Planning, Reasoning, Tool Use, and Memory. Building on this design space, we present a novel LLM agent search framework called AgentSquare, which introduces two core mechanisms, i.e., module evolution and recombination, to efficiently search for optimized LLM agents. To further accelerate the process, we design a performance predictor that uses in-context surrogate models to skip unpromising agent designs. Extensive experiments across six benchmarks, covering the diverse scenarios of web, embodied, tool use and game applications, show that AgentSquare substantially outperforms hand-crafted agents, achieving an average performance gain of 17.2% against best-known human designs. Moreover, AgentSquare can generate interpretable design insights, enabling a deeper understanding of agentic architecture and its impact on task performance. We believe that the modular design space and AgentSquare search framework offer a platform for fully exploiting the potential of prior successful designs and consolidate the collective efforts of research community. Code repo is available at https://github.com/tsinghua-fib-lab/AgentSquare.

Automatic diagnostic methods utilizing image analysis have seen a surge in development in recent years. However, achieving a comprehensive diagnosis necessitates considering clinical information. In the domain of skin disease, simply treating diagnosis as a multi-class image classification task poses inevitable bottlenecks. While existing methods incorporating Bayesian networks offer partial solutions, given the vast number of skin disease types, further integration with richer clinical information and diverse forms of deep neural networks would significantly escalate associated costs. This paper proposes an extensible diagnostic architecture. For newly added diseases and subsequently added clinical information, we integrate the output of deep neural networks into Bayesian networks more efficiently via semantic analysis. Through adjustments to node distribution types and the reduction of unnecessary connections, we expedite the development of a diagnostic system capable of addressing a wider range of skin diseases. Existing high-performing deep learning models, including multi-task and concept bottleneck, can be flexibly integrated into the system. Incorporating dermatologists’ expertise and reusing trained neural networks additionally reduces system development costs. At the same time, our architecture ensures excellent diagnostic accuracy. Through experiments leveraging the PAD-UFES-20 and SkinCon benchmarks, we have demonstrated that this architecture achieves a balanced diagnostic accuracy improvement of 19.3% over pure deep learning methods, with metrics including AUC surpassing those of existing published works. Our code can be accessed through the GitHub Repository: https://github.com/KevinInfinigon/MultimodalBayesianNetworks.

The consistency of pair-wise comparison matrix is a serious challenge for the multiple-criteria decision-making problem. However, existing methods are either too complicated to be applied in the revising process of the inconsistent comparison matrix or are difficult to preserve most of the original comparison information due to the use of a new pairwise comparison matrix. In this paper, a discrete region-based approach is proposed to improve the consistency of the pair-wise comparison matrix. When the decision makers feel confused or uncertain, they could express their evaluation as a discrete region containing multiple judgments, instead of a single result. A new data structure, named as set-matrix, is designed to store the combinations of those multiple judgments. An iterative searching algorithm is designed to find the pair-wise comparison matrix with approximate optimal consistency from the set-matrix. The experiments show that: 1) the consistency is significantly improved when the experts apply the discrete region evaluation instead of the single evaluation; and 2) the users can find the matrix with approximate optimal consistency quickly exploiting the iterative searching algorithm.

Numerous smart devices are deployed in smart grid for state measurement, decision-making and remote control. The security issues of smart devices attract more and more attention. In our work, the communication protocol, storage mechanism and authentication of smart devices are analyzed and a toolbox is developed to evaluate the security risks of smart devices. In this demo, our toolbox is applied to scan 3 smart meters/power monitor systems. A potential risk list is generated and the vulnerabilities are further verified.