Preprint
Aug 2026
MetaStrategy: Generative Ranking with Executable LLM Strategies
The method combines selection, relative-rank, and baseline-lift rewards, a self-competitive curriculum that feeds frequent strategies back as competitors, and Evaluator-routed reward-augmented on-policy distillation that transfers complementary 4B-parameter Teachers into a compact 0.8B-parameter Student.
Chengyu Lai, Jiuning Lin, Zhibo Xiao et al.
· 0 citations