Fig. 5: Metaprimitives are central to MPL’s performance.
From: Symbolic metaprogram search improves learning efficiency and explains rule learning in humans
A Follows (Fig. 3A), varying MPL's search steps per trial. B MPL's \({\log }_{e}\,-{\log }_{e}\) program prior (\({p}_{{{{\mathcal{P}}}}}(\widetilde{H})\)) relative to MPL's loge metaprogram prior (\({p}_{{{{\mathcal{M}}}}}(H)\)) for the highest-posterior hypotheses in each trial (dots; n = 1, 100 trials) with parity between the two priors (curve). C Follows (Fig. 3A) for the full MPL model and when lesioning the two priors.
