Chain of Thought: Matching Numeric Answers
Intermediate
The problem, from first principles
Chain-of-thought accuracy is measured by comparing the extracted answer with the gold answer. Numbers should match by value, so that '1,000' and '1000' count as the same answer.
From theory to code
Implement answers_match(pred, gold), a tolerant comparison of two answer strings.
Constraints
- Return a bool.
Hints
Hint 1
Normalize by removing commas and case; compare as floats if both parse, otherwise compare the strings.
Keywords
research-papersagentsreasoningprompting
Python
Run your code to see results
No results yet
Run your code to check it against the examples, or submit to run every test.