Search papers, labs, and topics across Lattice.
This paper investigates the impact of non-uniform parameter allocation across the layers of language models, revealing that earlier layers contribute more significantly to model performance than later layers. By implementing Tapered Language Models (TLMs), which allocate more capacity to earlier layers and less to later ones, the authors demonstrate improved perplexity and benchmark performance without increasing parameter count or computational cost. The results highlight the potential for depth-aware capacity allocation as a novel and effective design strategy in language model architecture.
Allocating more capacity to earlier layers in language models can significantly enhance performance, challenging the long-held uniform layer design paradigm.
Modern language models, including transformer, recurrent, and memory-based variants, share a common chassis: a stack of identical layers in which parameters are allocated uniformly across depth. This is a default inherited from the original transformer and largely unchanged since, yet a growing body of evidence suggests that layers contribute non-uniformly to the final output, with later layers refining the residual stream rather than transforming it. We ask whether parameter capacity should reflect this asymmetry. Our controlled experiment shows that, under a fixed budget, allocating more capacity to earlier layers and less to later layers improves perplexity over a uniform-width baseline, while the reverse allocation hurts. Building on this result, we introduce Tapered Language Models (TLMs), an architectural principle in which a parameter-bearing component is monotonically tapered across depth under a fixed total budget. MLPs are the natural site for this instantiation: they dominate parameter count across all modern LM families and expose width as a single, clean axis of variation. Across three model scales and four architectures (Transformer, Gated Attention, Hope-attention, and Titans), tapering MLP width via a smooth cosine schedule consistently improves perplexity and downstream benchmark performance over uniform baselines, at no additional parameter or compute cost. These findings establish depth-aware capacity allocation as a simple, architecture-agnostic axis of language model design, a free lever hidden in plain sight.
Mila