Quartet: Native FP4 Training Can Be Optimal for Large Language Models Paper • 2505.14669 • Published May 20, 2025 • 79
