Multiphysics simulation aims to predict and understand interactions between multiple physical phenomena, aiding in comprehending natural processes and guiding engineering design. The system of Partial Differential Equations (PDEs) is crucial for representing these physical fields, and solving these PDEs is fundamental to such simulations. However, current methods primarily yield numerical outputs, limiting interpretability and generalizability. We introduce T-NNGP, a hybrid genetic programming algorithm that integrates traditional numerical methods with deep learning to derive approximate symbolic expressions for multiple unknown functions within a system of PDEs. T-NNGP initially obtains numerical solutions using traditional methods, then generates candidate symbolic expressions via deep reinforcement learning, and finally optimizes these expressions using genetic programming. Furthermore, a universal decoupling strategy guides the search direction and addresses coupling problems, thereby accelerating the search process. Experimental results on three types of PDEs demonstrate that our method can reliably obtain human-understandable symbolic expressions that fit both the PDEs and the numerical solutions from traditional methods. This work advances multiphysics simulation by enhancing our ability to derive approximate symbolic solutions for PDEs, thereby improving our understanding of complex physical phenomena.

This paper introduces EmbodiedAgent, a hierarchical framework for heterogeneous multi-robot control. EmbodiedAgent addresses critical limitations of hallucination in impractical tasks. Our approach integrates a next-action prediction paradigm with a structured memory system to decompose tasks into executable robot skills while dynamically validating actions against environmental constraints. We present Mul-tiPlan+, a dataset of more than 18,000 annotated planning instances spanning 100 scenarios, including a subset of impractical cases to mitigate hallucination. To evaluate performance, we propose the Robot Planning Assessment Schema (RPAS), combining automated metrics with LLM-aided expert grading. Experiments demonstrate EmbodiedAgent’s superiority over state-of-the-art models, achieving 71.85% RPAS score. Real-world validation in an office service task highlights its ability to coordinate heterogeneous robots for long-horizon objectives.

Dark forest (DF) is the first successful decentralized real-time strategy (RTS) game. Based on smart contracts, the game explores the application possibilities of new technologies, and prefigures the operation of the social system in the metaverse era. In this study, we collect on-chain data from DF, map player behavior, and discuss players’ game asset management, combat strategies in extreme mode, and the game’s exit mechanism. We believe that blockchain games, such as DF, can be viewed as a new technological space with metaverse characteristics, in which individuals art free to challenge centralized social structures using decentralized technology while subjected to more secret postmodern social surveillance.

The notion of metaverse powered by blockchain technology has been buzzing in recent years, raising expectations for an advanced social vision of Web 3.0. Thus far, online multiplayer games are considered to be the sole counterparts of the full-state metaverse, and their in-game group behavior becomes a microscopic sample for estimating social dynamics in metaverse. In this paper, we select massive data from Dark Forest relating to players' behavior to explore how and to what extent the web 3.0 game simulates the social system. By analyzing and clustering the time series of players' behavior, We reveal the consistency of behavioral habits both inside and outside of the simulated gaming system; suggest that games based on blockchain technology have the advantage of digitally representing reality, and therefore constitute a key venue for future social practice.