An Open Recipe for IMO Gold: Training Nemotron for Olympiad Mathematics
We study how model post-training and test-time inference design affect natural-language proof generation for hard olympiad mathematics. Starting from Nemotron 3 Ultra, we train two specialist checkpoints using supervised fine-tuning and reinforcement learning, and evaluate checkpoint choice, verification, and refinement. Based on these findings, we present an open-model test-ti
Reading in Plain English mode — toggle below to switch
We study how model post-training and test-time inference design affect natural-language proof generation for hard olympiad mathematics. Starting from Nemotron 3 Ultra, we train two specialist checkpoints using supervised fine-tuning and reinforcement learning, and evaluate checkpoint choice, verification, and refinement. Based on these findings, we present an open-model test-ti
This work demonstrates a significant leap in using LLMs for complex mathematical reasoning, specifically achieving high scores on Olympiad-level problems. For ML engineers, it highlights advanced techniques in model post-training and inference design crucial for developing highly specialized and capable AI systems in challenging domains.