polca-stochastic-generative-optimization

Use language models as optimizers to iteratively improve complex systems (prompts, code, agent policies) through noisy feedback and sampling. POLCA maintains a priority queue of candidates, explores with epsilon-nets, and learns meta-insights from trial history.

Independent GitHub snapshot · Automated classification · Original documentation governs