The Depth Delusion: Why Transformers Should Be Wider, Not Deeper

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Fahim, Md Muhtasim Munif, Karim, Md Rezaul
Format: Preprint
Published: 2026
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866908795762900992
author Fahim, Md Muhtasim Munif
Karim, Md Rezaul
author_facet Fahim, Md Muhtasim Munif
Karim, Md Rezaul
contents Neural scaling laws describe how language model loss decreases with parameters and data, but treat architecture as interchangeable--a billion parameters could arise from a shallow-wide model (10 layers & 8,192 hidden dimension) or a deep-narrow one (80 layers & 2,048 hidden dimension). We propose architecture-conditioned scaling laws decomposing this dependence, finding that optimal depth scales as D* ~ C^0.12 while optimal width scales as W* ~ C^0.34, meaning width should grow 2.8x faster than depth. We discover a critical depth phenomenon: beyond D_crit ~ W^0.44 (sublinear in W), adding layers increases loss despite adding parameters--the Depth Delusion. Empirically, we validate these findings across 30 transformer architectures spanning 17M to 7B parameters, each trained on representative high-compute samples, achieving R^2 = 0.922. Our central finding: at 7B scale, a 64-layer model (6.38B params) underperforms a 32-layer model (6.86B params) by 0.12 nats, despite being significantly deeper. This demonstrates that optimal depth-width tradeoffs persist at the production scale.
format Preprint
id arxiv_https___arxiv_org_abs_2601_20994
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle The Depth Delusion: Why Transformers Should Be Wider, Not Deeper
Fahim, Md Muhtasim Munif
Karim, Md Rezaul
Machine Learning
Artificial Intelligence
Neural scaling laws describe how language model loss decreases with parameters and data, but treat architecture as interchangeable--a billion parameters could arise from a shallow-wide model (10 layers & 8,192 hidden dimension) or a deep-narrow one (80 layers & 2,048 hidden dimension). We propose architecture-conditioned scaling laws decomposing this dependence, finding that optimal depth scales as D* ~ C^0.12 while optimal width scales as W* ~ C^0.34, meaning width should grow 2.8x faster than depth. We discover a critical depth phenomenon: beyond D_crit ~ W^0.44 (sublinear in W), adding layers increases loss despite adding parameters--the Depth Delusion. Empirically, we validate these findings across 30 transformer architectures spanning 17M to 7B parameters, each trained on representative high-compute samples, achieving R^2 = 0.922. Our central finding: at 7B scale, a 64-layer model (6.38B params) underperforms a 32-layer model (6.86B params) by 0.12 nats, despite being significantly deeper. This demonstrates that optimal depth-width tradeoffs persist at the production scale.
title The Depth Delusion: Why Transformers Should Be Wider, Not Deeper
topic Machine Learning
Artificial Intelligence
url https://arxiv.org/abs/2601.20994