Can I read VAR-MATH: Evaluating True Math Reasoning on EtoBox?

VAR-MATH: Evaluating True Math Reasoning by Mariana is a document available to read on EtoBox.

What is VAR-MATH: Evaluating True Math Reasoning about?

The document introduces VAR-MATH, a symbolic evaluation framework aimed at assessing the true mathematical reasoning abilities of large language models (LLMs) by addressing issues of benchmark contamination and evaluation fragility. By transforming fixed numerical problems into symbolic templates and requiring models to solve multiple instances, VAR-MATH enhances evaluation robustness and reveals potential overfitting in existing reinforcement learning methods. Experimental results indicate significant perf

Author

Mariana

Language

EN