A new study reveals that cost-effective automated judging of natural-language mathematical proofs is possible using cheap open-weight models, achieving similar accuracy to frontier models at a significantly lower cost. This breakthrough has significant implications for the development and evaluation of math-reasoning systems. The study's findings could lead to more efficient and affordable methods for grading mathematical proofs.

OpenAI's unreleased Astra model has achieved a monumental breakthrough, solving ten previously intractable open problems in mathematics and theoretical computer science. This landmark event, detailed in a new publication, showcases AI's accelerating capability to drive fundamental scientific discovery, promising to reshape research methodologies and human-AI collaboration.