Safactory: A Scalable Agentic Infrastructure for Training Trustworthy Autonomous Intelligence
Xinquan Chen, Zhenyun Yin, Shan He, Bin Huang, Shanzhe Lei, Pengcheng Shi, Kun Cai, Bei Chen, Bangwei Liu, Zeyu Kang, Chao Huang, Yang Zhang, Wenjie Li, Ruijun Ge, Yajie Wang, Tianshun Fang, Tianyang Xu, Yiwen Cong, Meng Jin, Gaolei Li, Xuansheng Wu, Linhan Liu, Zijing He, An Li, Yan Teng, Xin Tan, Dongrui Liu, Jing Shao, ChaoChao Lu, Ji He, Jie Li, Chunfeng Song, Jinya Xu, Fan Song, Shujie Wang, Jianmin Qian, Jie Hou, Xuhong Wang, Yingchun Wang, Hui Wang, Xia Hu
Why It Matters
What makes this one worth your time
As AI systems evolve towards autonomy, a cohesive infrastructure like Safactory could streamline the development and evaluation of these systems, enhancing safety and reliability.
Safactory proposes a unified framework for training trustworthy autonomous agents.
Summary
The paper introduces Safactory, a scalable infrastructure designed to address the challenges of training trustworthy autonomous intelligence through a unified evolutionary pipeline that integrates simulation, data management, and reinforcement learning.
Key contributions
- Introduction of the Parallel Simulation Platform for trajectory generation.
- Development of the Trustworthy Data Platform for experience extraction.
- Creation of the Autonomous Evolution Platform for reinforcement learning.
Notable insights
- The integration of trajectory generation, storage, and evolution in a single framework is a non-trivial approach that could facilitate continuous improvement of autonomous agents.
- The focus on trustworthy intelligence highlights a growing concern in AI safety and ethics.
Possible limitations
- Not stated in the abstract.
Abstract
arXiv:2605.06230v2 Announce Type: replace Abstract: As large models evolve from conversational assistants into autonomous agents, challenges increasingly arise from long-horizon decision making, tool use, and real environment interaction. Existing agenticinfrastructure remain fragmented across evaluation, data management, and agent evolution, making it difficult to discover risks systematically and improve models in a continuous closed loop. In this report, we present \textbf{Safactory}, a scalable agent factory for trustworthy autonomous intelligence. Safactory integrates three tightly coupled platforms: a \textbf{Parallel Simulation Platform} for trajectory generation, a \textbf{Trustworthy Data Platform} for trajectory storage and experience extraction, and an \textbf{Autonomous Evolution Platform} for asynchronous reinforcement learning and on-policy distillation. As far as we know, Safactory is the first framework to propose a unified evolutionary pipeline for next-generation trustworthy autonomous intelligence.