Safactory: A Scalable Agent Factory for Trustworthy Autonomous Intelligence

2026-05-07

AI Summary

This work addresses the systemic risks and persistent optimization challenges arising from long-horizon decision-making, tool usage, and environmental interaction as large models evolve into autonomous agents. To this end, we propose Safactory—a unified evolutionary framework for trustworthy autonomous intelligence. Safactory tightly couples evaluation, data management, and model evolution through parallel simulation to generate interaction trajectories, a trusted data platform to distill experiential knowledge, and asynchronous reinforcement learning combined with online policy distillation to drive autonomous agent improvement. This framework establishes a scalable and continuously iterable infrastructure for agent development, significantly enhancing the capability to identify systemic risks and improving the efficiency of sustained model refinement.

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.

Problem

• autonomous agents
• long-horizon decision making
• trustworthy AI
• agent infrastructure
• systematic risk discovery

Innovation

• Scalable Agent Factory
• Trustworthy Autonomous Intelligence
• Parallel Simulation
• Autonomous Evolution
• Closed-loop Learning