X · @huggingface July 13, 2026 · X / Twitter

RT Sayak Paul: We show that best-of-N (BoN) is a much better baseline for diffusion test-time scaling when done right. Our core premise is that NFEs p…

RT Sayak PaulWe show that best-of-N (BoN) is a much better baseline for diffusion test-time scaling when done right.Our core premise is that NFEs provide a false sense of efficiency by ignoring several underlying factors, such as verifier costs.We then introduce Flash-BoN, which encourages broader exploration of sample