ImproveAnyTask: An Autonomous Post-Training Harness for Iterative Model Self-Improvement.

arXiv:2610.06347v1 Announce Type: new Abstract: Adapting general-purpose large language models to specific tasks requires substantial human effort in designing data and training strategies. Sustaining improvement is especially challenging because model updates