AI Paper Atlas

Research map / Architectures and efficiency / Pruning and efficient tuning

Shapeshifter: a Parameter-efficient Transformer using Factorized Reshaped Matrices

Aliakbar Panahi, Seyran Saeedi, Tom Arodz

NeurIPS 2021 · Architectures and efficiency › Pruning and efficient tuning · approach: Low-rank compression and merging · 6 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
replace Kronecker-based reshaped factorization with standard low-rank factorization (rank 16, embedding rank 256) (ablation)Shapeshifter · IWSLT'14 De-EnBLEU: 36.36 -> 26.68Hurt
replace Kronecker-based reshaped factorization with standard low-rank factorization (rank 16, embedding rank 256) (ablation)Shapeshifter · WMT'18 En-RoBLEU: 22.12 -> 17.47Hurt
replace Kronecker-based factorization with standard low-rank, embeddings not factorized (ablation)Shapeshifter 16/- · IWSLT'14 De-EnBLEU: 35.85 -> 32.01Hurt
replace Kronecker-based factorization with standard low-rank, embeddings not factorized (ablation)Shapeshifter 16/- · WMT'18 En-RoBLEU: 21.93 -> 19.34Hurt
replace Kronecker-based factorization with standard low-rank (rank 16/256) (ablation)Shapeshifter · IWSLT'14 De-Entrain loss: 2.59 -> 2.97Hurt
replace Kronecker-based factorization with standard low-rank (rank 16/256) (ablation)Shapeshifter · WMT'18 En-Rotrain loss: 2.19 -> 2.32Hurt
replace Kronecker-based factorization with standard low-rank (rank 16/256) (ablation)Shapeshifter · IWSLT'14 De-Endev loss: 2.67 -> 2.99Hurt
replace Kronecker-based factorization with standard low-rank (rank 16/256) (ablation)Shapeshifter · WMT'18 En-Rodev loss: 3.09 -> 3.37Hurt
replace PHM layers (n=16) with Shapeshifter 16/256 (about four times fewer parameters)T5-small Transformer · WMT'18 En-RoBLEU: 22.38 -> 22.12Mixed
replace PHM layers (n=16) with Shapeshifter 16/256T5-small Transformer · IWSLT'14 De-EnBLEU: 36.49 -> 36.36Mixed
compress Transformer with Shapeshifter 16/256 versus uncompressedT5-small Transformer · WMT'18 En-RoBLEU: 24.30 -> 22.12Hurt
increase Shapeshifter rank from 16/256 to 24/256 (ablation)Shapeshifter · WMT'18 En-RoBLEU: 22.12 -> 22.56Improved

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.

Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.

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