AI Paper Atlas

Research map / Architectures and efficiency / Attention and transformers

The geometry of hidden representations of large transformer models

Lucrezia Valeriani, Diego Doimo, Francesca Cuturello, Alessandro Laio, Alessio ansuini, Alberto Cazzaniga

NeurIPS 2023 · Architectures and efficiency › Attention and transformers · approach: Transformer block and depth design · 16 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
use plateau-layer (low ID) representations instead of the last hidden layer for nearest-neighbor remote homologyESM-2 · SCOPe remote homologyneighbor overlap with superfamily labels: 0.4 -> 0.8Improved
use intermediate layer at ID minimum instead of final layeriGPT · ImageNet 300-class labelsoverlap with ground truth labels: peak chi_gt 0.15 (S), 0.27 (M), 0.35 (L)Improved
increase model sizeiGPT · ImageNetpeak overlap with labels: 0.15 -> 0.35Improved
increase model sizeESM-2 · ProteinNetpeak intrinsic dimension: 20 -> 32 at peak; plateau stays 5-7Mixed
extract representations after attention maps or MLP instead of after first norm layer (ablation)iGPT-L · ImageNetID and overlap profilesNo clear effect
vary neighborhood size k from 1 to 50 (ablation)overlap analysis · ESM-2 650M and iGPT-Loverlap profile shapeNo 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