Library of Layers: Reworking Transformer Layer Order at Inference Time

A multi-institution team proposed Library of Layers (LoL), an architecture that abandons fixed Transformer layer ordering and instead routes each input through a shared library of expert blocks at inference time, enabling dynamic—and controllable—computation. The paper gained traction on Reddit after researcher Sean Mcleish introduced it.

2026-11-03 ~ 2026-11-04 · 3 related posts

• Library of Layers: EPFL's architecture reorders transformer layers at inference — Abdellah Rahmani · 2026-11-03
• Library of Layers: models route transformer block order per input to control algorithms — jonasgeiping · 2026-11-04
• Library of Layers (LoL): Paper Proposes Adaptively Ordering Transformer Layers at Runtime — Thrumpwart · 2026-11-04