<2-3 sentences summarising what the project addresses and solves, the target audience, and the benefit.> SimWorld is an open-ended simulation platform for developing and evaluating LLM/VLM AI agents in complex physical and social environments. It offers a realistic, Unreal Engine 5-based world with accurate physics and language-driven procedural generation, enabling systematic training and evaluation of advanced agent behaviors like navigation, planning, and cooperation. The platform targets AI researchers and developers seeking a robust environment for embodied AI and agent research.

How It Works

The architecture comprises three layers: the Unreal Engine Backend for realistic environments and physics, an Environment layer supporting procedural city generation and gym-like APIs for agents, and an Agent layer enabling LLM/VLM reasoning over multimodal observations and natural language actions via an action planner. This modular design facilitates dynamic world generation, agent control, and performance benchmarking, integrating LLM/VLM capabilities with embodied interaction.

Quick Start & Requirements

Requires Python 3.10+ and Conda. Install the Python client via pip install -e . after cloning the repository. Download the SimWorld server executable (v0.1.0) for Windows or Linux. Examples are provided in scripts/ and a minimal demo is available at examples/minimal_demo.ipynb.

Highlighted Details

• Accepted as a spotlight paper at NeurIPS 2025.
• Demo accepted by CVPR 2025 Demonstration Track.
• White paper available on arXiv.
• Features realistic physics, language-based procedural generation, and multi-modal perception for agents.

Maintenance & Community

The README does not detail specific contributors, community channels (e.g., Discord, Slack), or a public roadmap.

Licensing & Compatibility

The repository's license is not explicitly stated in the README, which may pose compatibility concerns for commercial use or integration into closed-source projects.

Limitations & Caveats

The SimWorld server is at version 0.1.0, indicating an early-stage release. The README does not specify known bugs, unsupported platforms, or missing features, but the early version suggests potential for ongoing development and changes.