Do Depth-Grown Models Overcome the Curse of Depth? An In-Depth Analysis
This work addresses inefficiencies in deep Transformer training for AI researchers, offering incremental improvements in model design and performance.
The paper tackles the problem of limited depth utilization in Transformers, known as the Curse of Depth, by analyzing depth-grown models and showing that gradual middle stacking improves depth utilization and alters residual stream structure, with a proposed modification yielding further gains in reasoning benchmarks.
Gradually growing the depth of Transformers during training can not only reduce training cost but also lead to improved reasoning performance, as shown by MIDAS (Saunshi et al., 2024). Thus far, however, a mechanistic understanding of these gains has been missing. In this work, we establish a connection to recent work showing that layers in the second half of non-grown, pre-layernorm Transformers contribute much less to the final output distribution than those in the first half - also known as the Curse of Depth (Sun et al., 2025, Csordás et al., 2025). Using depth-wise analyses, we demonstrate that growth via gradual middle stacking yields more effective utilization of model depth, alters the residual stream structure, and facilitates the formation of permutable computational blocks. In addition, we propose a lightweight modification of MIDAS that yields further improvements in downstream reasoning benchmarks. Overall, this work highlights how the gradual growth of model depth can lead to the formation of distinct computational circuits and overcome the limited depth utilization seen in standard non-grown models.