Independent & Equitable Evaluation

IMProofBench is committed to providing independent, equitable, and transparent evaluations of AI mathematical reasoning capabilities. Financial support from any organization does not influence our scientific evaluation methodology, which applies the same blinded, model-agnostic protocols to all AI systems.

Evaluations of models from companies that have provided financial support follow the same rigorous standards as all other evaluations.

Data Protection Policy

We do not share question or solution data with third parties, except in the following controlled circumstances:

• Peer Review: Submitted questions are reviewed by expert mathematicians to ensure quality and appropriate difficulty.
• API Evaluations: Questions are evaluated via official AI provider APIs, protected by data privacy guarantees and zero-data-retention policies where available.
• Public Sample Questions: A small set of sample questions are released publicly, only with explicit consent from their authors.