←Back to NewsAI News/LlmspaperLlmsTraining InfraTsinghua's SMELT Cuts AI Training Costs 18% by Looping Layers TwiceNew scaling laws show that looping the middle half of a Mixture-of-Experts model twice saves up to 18% of training compute at matched budgets.SourceAlphaSignalPublishedSep 1, 2026, 2:52 PMAuthorAlphaSignal NewsroomRead1 min readNew scaling laws show that looping the middle half of a Mixture-of-Experts model twice saves up to 18% of training compute at matched budgets.Reporting is indexed from AlphaSignal. Rights remain with the original publisher and cited sources.Read original report ↗Next readsMagic · newsMagic Matches DeepSeek V4 Pro Base Using 50x Less ComputeHark · newsFigure AI's Brett Adcock Bets Gigawatt NVIDIA Deal on Hark HandoffAlphaSignal · paperOne Hidden Ratio Controls How Fast AI Models Actually Learn
