Researchers present Argus, a persistent, self-evolving agentic runtime designed for long-horizon reasoning that separates stable user intent from operational objectives. The system utilizes Manager, Planner, Engineer, and Reviewer roles to execute bounded missions over durable project state, allowing for autonomous execution between operator-owned escalation points.
- Across seven GPT-5.5 benchmark arenas, Argus achieves about 78% on SWE-Bench Pro compared to 59% for Direct Copilot, using 1.41 times the aggregate tokens.
- After verification-gated self-evolution, mature SWE-Bench waves use 21% fewer solve-input tokens and 15% less active workflow time per task than startup waves.
- The system records 34 verifier recoveries and 22 strict review-loop rescues, reaching 76.8% on AARRI-Bench and a 28.0-point gap on mathematical data synthesis.
- Beyond benchmarks, an optimized RWKV6 kernel was merged upstream, and six paper pipelines completed 254 missions with 16 stage rollbacks.
These results demonstrate that a fixed-weight, self-evolving harness can revise, recover, and accumulate verified approaches while producing structured trajectories for future supervised and reinforcement learning.