Mitigating Safety Tax via Distribution-Grounded Refinement in Large Reasoning Models Paper • 2602.02136 • Published Feb 2 • 7
