benchmarks/procedural/evaluators/numeric.py

import re
from typing import Any
from benchmarks.procedural.interfaces import Evaluator

class NumericEvaluator(Evaluator):
    @property
    def name(self) -> str:
        return "numeric"
        
    def evaluate(self, answer: str, ground_truth: Any) -> bool:
        if not isinstance(ground_truth, (int, float)):
            try:
                ground_truth = float(ground_truth)
            except ValueError:
                return False
                
        # Extract the last number from the model's answer (common heuristic)
        # Handle negative numbers and decimals
        matches = re.findall(r'-?\d+\.?\d*', str(answer))
        
        if not matches:
            return False
            
        model_val = float(matches[-1])
        
        # Check tolerance (e.g. 1e-4)
        return abs(model_val - float(ground_truth)) < 1e-4