Problem
Models often write Greek as Unicode (θ, α, φ) while gold uses LaTeX (\theta, \alpha, \phi). Pi already folds (\pi/π → pi), but other letters do not:
math_equal(r"2\theta", "2θ") # False
math_equal(r"\alpha", "α") # False
normalize_math_answer(r"2\theta") # "2\theta"
normalize_math_answer("2θ") # "2θ"Expected
Common Greek letters normalize to the same ascii token (like pi) so LaTeX and Unicode spellings grade equal.
Environment
Problem
Models often write Greek as Unicode (
θ,α,φ) while gold uses LaTeX (\theta,\alpha,\phi). Pi already folds (\pi/π→pi), but other letters do not:Expected
Common Greek letters normalize to the same ascii token (like
pi) so LaTeX and Unicode spellings grade equal.Environment