AIMO Interpretability Challenge
The AIMO Interpretability Challenge tests if interpretability methods can distinguish robust mathematical reasoning from brittle shortcuts in frontier language models. By subjecting models to structured problem variations, participants create systems that classify whether a model's success generalizes or is fragile, aiming to establish a reliable foundation for assessing model reliability.