Update README.md

Rename benchmark split to test

Remove stray Z from model scores image

Add separate dataset configs for numbers and results

Add number submission form link

Restore image-matching 32 model results

Use fixed results with missing legacy models merged in

Restore full model results and update results table

Add confusion counts and bias notes to results table

Add model results table to dataset card

Upload HumanNumberEval dataset card and benchmark files

initial commit