(phase1)[experiment/question]: does Paco outperform mothra-yolo

per character-word pulled vs actual, per neume-vs-actual etc for each branch, then compare against mothra phase1v1 models and then the mothra phase1v2 models @homework36 is training

low priority until summer wrapup is complete

Start by locating the evaluation outputs or scripts for each branch, then identify the mothra phase1v1 models and the phase1v2 models being trained by @homework36. Compare per-character-word and per-neume pulled-versus-actual results across the branches and models; done means a clear comparison of whether Paco outperforms mothra.

Written by the indexing model from the issue text.