On the Non-decoupling of Supervised Fine-tuning and Reinforcement Learning in Post-training Paper • 2601.07389 • Published 3 days ago • 1