1 paper · 1 filter
Kalle Kujanpää, Kalle Kujanpää, Ning Liu +5
Production LLM agents often waste latency and reliability by regenerating code for the same procedural steps on every request. We replace this inference-time coding loop with an ag…