Apply suggestions from code review

Co-authored-by: Clémentine Fourrier <[email protected]>

Author: hynky1999

src/lighteval/metrics/utils/extractive_match_utils.py

@@ -151,9 +151,9 @@ def lazy_latex_regex(latex_config: LatexExtractionConfig, language: Language) ->
         r"("
         r"(?<!\\)\$\$(?P<latexDisplayDollar>[\s\S]+?)(?<!\\)\$\$|"  # $$...$$ (display math, can be multiline)
         r"(?<!\\)\\\[(?P<latexDisplayBracket>[\s\S]+?)(?<!\\)\\\]|"  # \[...\] (display math, can be multiline)
-        r"(?<!\\|\d)\$(?P<latexInlineDollar>(?:\\[$]|[^\n$])+?)(?<!\\)\$|"  # $...$ (inline math, single line, allows escaped $), we make sure it's not preceed by a digit to minimize false positives with actualy dollar unit
+        r"(?<!\\|\d)\$(?P<latexInlineDollar>(?:\\[$]|[^\n$])+?)(?<!\\)\$|" # $...$ (inline math, single line, allows escaped $), we make sure it's not preceded by a digit to minimize false positives containing dollar as a unit
         r"(?<!\\)\\\((?P<latexInlineParenthesis>[^\n]+?)(?<!\\)\\\)|"  # \(...\) (inline math, single line)
-        r"(?<!\\)\[(?P<latexInlineBracket>[^\n$]+?)(?<!\\)\]"  # [....] While this is no a valid display math llms like to generate it, allow it
+                r"(?<!\\)\[(?P<latexInlineBracket>[^\n$]+?)(?<!\\)\]"  # [....] While this is not a valid display, math LLMs like to generate it. We allow it
         rf"){percent_re_group}?"
     )
 
@@ -306,7 +306,7 @@ def extract_expr(match: re.Match) -> tuple[str | sympy.Expr | None, str]:
     is_percentage = True if groups.get("percent", None) else False
 
     if integer or decimal:
-        # This makes sure we can convert numebrs like 0001 to 1, do note that this can convert 0 to ''
+        # This makes sure we can convert numbers like 0001 to 1. Do note that this can convert 0 to '', so we assume an empty string was 0 and convert it back afterwards.
         integer = integer.translate(str.maketrans("", "", ", ")).lstrip("0")
         if len(integer) == 0:
             integer = "0"

src/lighteval/metrics/utils/math_comparisson.py

@@ -336,9 +336,9 @@ def sympy_expr_eq(gold: Basic | MatrixBase, pred: Basic | MatrixBase, precision:
     Returns:
         True if expressions are equal by any comparison method, False otherwise
     """
-    # If refernce is relational, but target it's not it's possible it's case of k=x+1+z, so we just take x+1+z
-    # We assume that the gold never contains symplifications, so we don't handle that case
-    # e.g 1+1+1=3 will never be simplified to 3, it would be possibly by doing lhs-rhs == 0, but ehhh just make the gold simpler
+    # If the reference is relational, but the target is not, it's possible it's a case of answer=x+1+z, so we just take x+1+z
+    # We assume that the gold never needs to be simplified, so we don't handle that case
+    # e.g 1+1+1=3 will never be simplified to 3; it would be possible to do so with lhs-rhs == 0, but we assume the gold is at its most simplified form.
     # The new latex2sympy2 will actually convert such cases automatically, but so this is in theory not needed
     if isinstance(gold, Eq) and not isinstance(pred, Relational) and isinstance(gold.lhs, Symbol):
         gold = gold.rhs