Frontier language models improve with additional test-time computation, but serial reasoning or uncoordinated parallel sampling can be compute-inefficient under fixed inference budgets. We propose SELFCEST, which equips a base model with the ability to spawn same-weight clones in separate parallel contexts by agentic reinforcement learning. Training is end-to-end under a global task reward with shared-parameter rollouts, yielding a learned controller that allocates both generation and context budget across branches. Across challenging math reasoning benchmarks and long-context multi-hop QA, SELFCEST improves the accuracy-cost Pareto frontier relative to monolithic baselines at matched inference budget, and exhibits out-of-distribution generalization in both domains.
@article{arxiv.2602.13262,
title = {General learned delegation by clones},
author = {Darren Li and Meiqi Chen and Chenze Shao and Fandong Meng and Jie Zhou},
journal= {arXiv preprint arXiv:2602.13262},
year = {2026}
}
Comments
Code available at https://github.com/SuffixAutomata/SELFCEST