AI Paper Atlas

Research map / Probabilistic and scientific modeling / Molecules and proteins

Learning Sparse Prototypes for Text Generation

Junxian He, Taylor Berg-Kirkpatrick, Graham Neubig

NeurIPS 2020 · Probabilistic and scientific modeling › Molecules and proteins · approach: Neuro-symbolic probabilistic layers · 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 the learned sparse prototype model instead of the NLM baselineLanguage modeling · MSCOCOPPL: 20.0 -> 18.6Improved
Use the learned sparse prototype model instead of the NLM baselineLanguage modeling · Yelp MediumPPL: 74.7 -> 61.9Improved
Use the learned sparse prototype model instead of the NLM baselineLanguage modeling · Yelp LargePPL: 34.2 -> 30.2Improved
Interpolate the sparse model with an NLMSparse Neural Editor · Yelp LargePPL: 30.2 -> 20.2Improved
Use learned prototypes instead of heuristic pre-clusteringNeural Editor with 778 prototypes · MSCOCOPPL: 19.5 -> 18.6Improved
Use learned prototypes instead of heuristic pre-clusteringNeural Editor with 778 prototypes · MSCOCOPOS-BLEU: 40.4 -> 42.5Improved
Increase Dirichlet concentration α from 10⁻³ to 0.1 (ablation)Sparse Neural Editor · MSCOCOPPL: 18.9 -> 18.6Improved
Increase Dirichlet concentration α from 0.1 to 0.2 (ablation)Sparse Neural Editor · MSCOCOPPL: 18.6 -> 19.0Hurt
Increase Dirichlet concentration α from 0.2 to 0.3 (ablation)Sparse Neural Editor · MSCOCO# prototypes: 16K (40%) -> 22K (56%)Mixed
Increase Dirichlet concentration α from 10⁻³ to 0.5 (ablation)Sparse Neural Editor · Yelp MediumPPL: 63.6 -> 61.9Improved
Increase Dirichlet concentration α from 0.5 to 10 (ablation)Sparse Neural Editor · Yelp MediumPPL: 61.9 -> 63.2Hurt
Increase Dirichlet concentration α from 0.7 to 10 (ablation)Sparse Neural Editor · Yelp LargePPL: 30.2 -> 30.3Hurt

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