AI Paper Atlas

Research map / Architectures and efficiency / Attention and transformers

Signal Propagation in Transformers: Theoretical Perspectives and the Role of Rank Collapse

Lorenzo Noci, Sotiris Anagnostidis, Luca Biggio, Antonio Orvieto, Sidak Pal Singh, Aurelien Lucchi

NeurIPS 2022 · Architectures and efficiency › Attention and transformers · approach: Transformer expressivity and learnability · 4 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
increase token alignment toward rank collapseTransformer self-attention · Transformer initializationquery/key gradient normHurt
scale residual branches with depth as α₁², α₂² = O(1/L)Transformer · Infinite-depth signal propagationtoken correlationImproved
use nonzero residual strengths that are constant with depth instead of depth-dependent scaling (ablation)Transformer · Infinite-depth signal propagationtoken correlationHurt
remove depth-dependent residual scaling (ablation)Transformer · Finite-depth signal propagationtoken correlationHurt
add layer normalization (ablation)Transformer with residual scaling · Initializationtoken correlationNo clear effect
use PRE-LN instead of no layer normalization (ablation)Transformer without residual scaling · Initializationtoken correlationImproved
standardize input variance with layer normalizationTransformer self-attention · Initializationquery/key gradient normImproved
train with residual scaling and inverse-temperature parameters using SGD instead of AdamTransformer encoder-decoder · IWSLT14 German-to-EnglishBLEUMixed
increase input varianceTransformer self-attention · Initializationquery/key gradient normImproved
increase input varianceTransformer self-attention · Initializationvalue gradient normImproved

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy