Red queen hypothesis – A new way forward for self-improving AI

Researchers have proposed the 'Red Queen' hypothesis for AI, where self-improving agents and their evaluators evolve simultaneously to prevent performance plateaus. This method allows agents to continue improving beyond the limitations of fixed benchmarks.