Data as of Sep 18, 2026 · Based on 3,315,446 AI responses across 10,525 prompts · See how Parse measures this
AgentMercuryminstar.github.io · Evidence updatingAgentMercury synthesizes executable, verifiable business environments from high-level scenarios so reinforcement-learning policies can train inside scalable, task-agnostic worlds. Using Planet, a scenario is turned into a complete world with services, tools, state schemas, seeded data, and executable invariants, enabling one world to support many tasks with deterministic post-episode grading. The system spans 4,783 worlds across 14 industries and 50 countries, with 43,300 verifiable RL tasks, and experiments show agents trained inside these worlds achieve rising rewards and transfer to established benchmarks.
Parse Score
