TinyAstral: Transferring a self-evolving harness to small generalist manipulation agents

TinyAstral proposes a two-stage method: first, freeze an open-weight teacher model and use recursive self-improvement (RSI) to have frontier LLM agents read the teacher's trajectories and propose harness modifications, with each modification accepted only if it passes a paired statistical test; then freeze the final harness and distill a small student model.