Feyospace enables strong cyber agents trained by small expert team
Feyospace-v1: How the Cyber Mercury Seven Trained Frontier Cyber Models
Artificial Intelligence
Summary
Training smart computer agents to solve complex cybersecurity tasks is usually expensive due to the need for realistic practice environments and good guidance. The authors created a system called Feyospace that combines five tools and a special data engine to make training efficient and reliable. Their approach produces models that perform much better on cybersecurity challenges and ranks highly on a global leaderboard. They also show that a small team of seven people can train powerful open-source cyber agents with these methods.
What this means in practice
- •For cybersecurity teams: Deploy trained cyber agents to automate detection and exploitation of vulnerabilities in realistic test environments.
- •For security software developers: Integrate the Feyospace training framework to build more capable open-weight cyber defense and attack models efficiently.$Commercial implications: Enables creating advanced security tools powered by trained cyber agents for commercial cybersecurity products.
Authors
Zongjie Li, Alan Z. W, John Nicolas J, Walter H. F, Scott Donald L, Gordon Y. P, Deke X
Abstract
Training capable cyber agents is often treated primarily as a problem of model scale, yet open-weight post-training is constrained more directly by the cost of executable environments, reliable multi-turn supervision, and access to strong teachers. We present a data-centric framework that addresses these bottlenecks through five complementary systems: Choulea analyzes hidden reasoning signatures, SkyReal reduces teacher-sampling cost, Hongzwang bypasses API restrictions on teacher execution, PSBreakup restores capabilities weakened by model merging, and Kreator converts expert interventions into trainable reasoning. Our data engine constructs resettable coding, vulnerability, CTF, kernel-history, full-exploit, firmware, and device-backed environments. Candidate trajectories are retained only after execution verification and evidence auditing, yielding 164,269 trajectories for long-context supervised fine-tuning. The three checkpoints improve over their starting models by an average of 23.76% on the full CyberGym suite and 10.49% across the pooled CTF suites. As of September 1, 2026, Feyospace-s1 achieves a verified success rate of 63.24% and ranks 10th on the official CyberGym leaderboard, while all three checkpoints rank 1st among models at comparable parameter scales. To our knowledge, this is the first end-to-end demonstration that a seven-person independent team can train open-weight models with leading agentic cyber capability.