EigenData
A self-evolving, multi-agent engine for AEI that generates, verifies, and refines complex datasets across function-calling, agents, and multimodal tasks.
- Hierarchical agents: Orchestrator/verifier on top; workers for task proposal & decomposition, user simulation, trajectory generation, and trajectory→dataset conversion.
- Self-evolving loops: Compose → self-verify each round, with optional human feedback to steer the next iteration.
- Rich correctness signals: Checkers for tool calls, agent state, and multimodal outputs produce structured pass/fail and scores.