Meta-Harness: End-to-End Optimization of Model Harnesses

ArXi:2603.28052v1 Announce Type: new The performance of large language model (LLM) systems depends not only on model weights, but also on their harness: the code that determines what information to, retrieve, and present to the model. Yet harnesses are still designed largely by hand, and existing text optimizers are poorly matched to this setting because they compress feedback too aggressively. We