method · first seen 30 Oct 2026
hierarchical residual
Also: HiRes
Decoupling residual stream width from per-layer compute width by grouping layers into blocks with a wider outer residual.
Mentions per week
PapersCodeDiscussion
Earliest sources
- Hierarchical Residuals: A Minimalist Decoupling of Residual Width from Compute WidthPapers · 30 Oct 2026 · Ziteng Wang, Jianqiao Lu, Xinyu Ma, Jintao Zhang, Jun Zhu, Yiyuan Ma, Jianfei Ch…
Latest sources
- Hierarchical Residuals: A Minimalist Decoupling of Residual Width from Compute WidthPapers · 30 Oct 2026 · Ziteng Wang, Jianqiao Lu, Xinyu Ma, Jintao Zhang, Jun Zhu, Yiyuan Ma, Jianfei Ch…