Basu, M. (2025). DeepSeek’s self-correcting AI model aces tough maths proofs. Nature. https://doi.org/10.1038/d41586-025-03959-9