Testing AI models that learn by reflecting on their own mistakes

Original: Going recursive (part I): Applying RLM-GEPA to AppWorld 🌎

Source: x.com ↗