Reasoning over mathematical objects: on-policy reward modeling and test time aggregation

Open paper

Benchmark	Model	Rank	Results
mathematical-reasoning-on-aime-2025	Principia-4B	–	Accuracy: 78.12
mathematical-reasoning-on-aime24	Principia-4B	–	Acc: 81.06
multiple-choice-question-answering-mcqa-on-gpqa-diamond	Principia-4B	–	Accuracy: 63.83