AI Paper Atlas

Research map / Architectures and efficiency / Attention and transformers

Language Through a Prism: A Spectral Approach for Multiscale Language Representations

Alex Tamkin, Dan Jurafsky, Noah Goodman

NeurIPS 2020 · Architectures and efficiency › Attention and transformers · approach: Language model pretraining and induction · 13 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 a prism layer after the last BERT layerBERT · 20 Newsgroupsaccuracy: 32.21 -> 51.01Improved
Add a prism layer after the last BERT layerBERT · Switchboard Dialog Speech Actsaccuracy: 47.09 -> 54.02Improved
Add a prism layer after the last BERT layerBERT · Penn Treebankaccuracy: 95.86 -> 94.41Hurt
Use only the lowest-frequency sector (ablation)BERT + Prism · 20 Newsgroupsaccuracy: 45.1% versus 51.0% for the full modelHurt
Use only the highest-frequency sector (ablation)BERT + Prism · 20 Newsgroupsaccuracy: 5.3% versus 51.0% for the full modelHurt
Use only the highest-frequency sector (ablation)BERT + Prism · Penn Treebankaccuracy: 84.1% versus 94.4% for the full modelHurt
Use only the lowest-frequency sector (ablation)BERT + Prism · Penn Treebankaccuracy: 16.8% versus 94.4% for the full modelHurt
Place prism layers after every BERT layer instead of only after the last layer (ablation)BERT + Prism · 20 Newsgroupsaccuracy: 45.2% (-5.8%)Hurt
Place prism layers after every BERT layer instead of only after the last layer (ablation)BERT + Prism · Switchboard Dialog Speech Actsaccuracy: 51.8% (-2.2%)Hurt
Place prism layers after every BERT layer instead of only after the last layer (ablation)BERT + Prism · Penn Treebankaccuracy: 94.0% (-0.4%)Hurt
Use a low-pass spectral filterPretrained BERT representations · 20 Newsgroupsprobing accuracyImproved
Use a high-pass spectral filterPretrained BERT representations · Penn Treebankprobing accuracyMixed

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