73. An approach that uses RL-trained small-scale models as sub-agents while dedicating the frontier model to judgment… by @kuto_bopro (kuto) · backlist 2026-05-08 · rubric 88.0