AI Paper Atlas

Research map / Architectures and efficiency / Attention and transformers

The Structural Origin of Attention Sink: Variance Discrepancy, Super Neurons, and Dimension Disparity

Siquan Li, Kaiqi Jiang, Jiacheng Sun, Tianyang Hu

ICML 2026 · Architectures and efficiency › Attention and transformers · approach: Sparse and structured attention · 0 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
add head-wise RMSNorm after value aggregationLlama-2 152M softmax attention · OpenWebText pretrainingValidation loss: 2.7812 ± 0.0109 -> 2.7421 ± 0.0066Improved
add head-wise RMSNorm after value aggregationLlama-2 152M softmax attention · OpenWebText pretrainingTrain loss: 2.7483 ± 0.0118 -> 2.7073 ± 0.0095Improved
add head-wise RMSNorm after value aggregationLlama-2 152M softmax attention · OpenWebText pretrainingEffective rank: 343.71 ± 15.63 -> 445.96 ± 5.37Improved
add head-wise RMSNorm after value aggregationLlama-2 152M softmax attention · OpenWebText pretrainingDimension disparity: 82.67 ± 8.09 -> 33.74 ± 2.73Improved
replace softmax with unnormalized sigmoid attention (ablation)Llama-2 152M · OpenWebText pretrainingValidation loss / convergenceHurt
replace softmax with sigmoid attention (ablation)Llama-2 152M · OpenWebText pretrainingAttention sinkImproved
block token 10 from attending to earlier tokens (mask) (ablation)Llama-2-7B · WikiText-2Attention to tokenImproved
amplify variance of token 10 by lambda (ablation)Llama-2-7B · random tokensAttention receivedImproved
scale norm by lambda instead of variance (control) (ablation)Llama-2-7B · random tokensAttention receivedNo clear effect

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