Looped transformers: more looping boosts reasoning but with diminishing returns past 3 passes and fewer CoT tokens

nrehiew_ · x · 2026-11-21

nrehiew summarizes findings from the TML-d24-varloop-research model (variable looped layers):

Related event: Looping benefits taper after 3 iterations while inference tokens drop, study finds(2 posts)→

Original post →

More from Models

Models channel →