Papers
Topics
Authors
Recent
Gemini 2.5 Flash
Gemini 2.5 Flash
123 tokens/sec
GPT-4o
11 tokens/sec
Gemini 2.5 Pro Pro
47 tokens/sec
o3 Pro
5 tokens/sec
GPT-4.1 Pro
4 tokens/sec
DeepSeek R1 via Azure Pro
33 tokens/sec
2000 character limit reached

A Random Matrix Theory Perspective on the Learning Dynamics of Multi-head Latent Attention (2507.09394v1)

Published 12 Jul 2025 in cs.LG

Abstract: In this work, we study how multi-head latent attention (MLA), a popular strategy for compressing key/value memory, affects a transformer's internal capacity during pretraining. Using a lightweight suite of Marchenko-Pastur (MP) diagnostics, we analyze the spectrum of the $W_{Q}W_{K}\top$ gram matrix throughout training, comparing three variants: the standard multi-head attention (MHA) baseline, MLA-PreRoPE with rotary applied before compression, and MLA-Decoupled, which shares a single rotary sub-vector across all heads. Our random matrix analysis reveals \textbf{three key findings:} \textbf{ i)} capacity bottlenecks emerge locally: both MHA and MLA-PreRoPE exhibit sharp, early spikes in specific layers that persist and propagate, disrupting the balance between bulk and outlier directions; \textbf{ ii)} these spikes coincide with rank collapse, concentrating the model's expressivity into narrow subspaces; \textbf{ iii)} only the decoupled variant prevents this cascade, maintaining broad spectral support and suppressing outlier formation across layers. These results underscore that \emph{how} rotary embeddings are applied is just as critical as \emph{where} compression occurs. Sharing rotary components across heads mitigates spectral fragmentation and preserves representational capacity.

Summary

We haven't generated a summary for this paper yet.