Red Queen Gödel Machine coevolves agents and evaluators
A paper proposes including evaluators in recursive self-improvement of agents, rather than assuming that verifiers, benchmarks, or labeled datasets remain fixed.
The paper proposes coevolving agents and evaluators during recursive self-improvement. It questions methods that depend on a fixed verifier, benchmark, or labeled dataset, which may no longer be suitable as agents improve.
For engineers building self-improving agents, the work highlights a practical issue: evaluation criteria may also need to evolve. If you use AI to study or apply this idea, avoid entering personal data or confidential information without authorization, and follow your organization’s security rules.