AI Paper Atlas

Research map / Architectures and efficiency / Attention and transformers

Compressive Transformers for Long-Range Sequence Modelling

Jack W. Rae, Anna Potapenko, Siddhant M. Jayakumar, Chloe Hillier, Timothy P. Lillicrap

ICLR 2020 · Architectures and efficiency › Attention and transformers · approach: Sparse efficient transformers for long context · 48 citations (OpenAlex)

Read the paper

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
compress old memories into a secondary memoryTransformerXL-style long-context language model · PG-19 testperplexity: 36.3 -> 33.6Improved
compress old memories into a secondary memory24-layer TransformerXL · Enwik8BPC: 0.98 -> 0.97Improved
compress old memories into a secondary memory18-layer TransformerXL · WikiText-103 testperplexity: 18.1 -> 17.1Improved
use convolutional compression with attention-reconstruction loss instead of convolution with BPTT (ablation)Compressive Transformer · Enwik8BPC: 0.996 -> 0.973Improved
use attention-reconstruction loss rather than auto-encoding loss for convolutional compression (ablation)Compressive Transformer · Enwik8BPC: 0.984 -> 0.973Improved
use convolutional compression rather than max pooling (ablation)Compressive Transformer · Enwik8BPC: 0.986 -> 0.973Improved
increase compressed memory size from 512 to 3072 (ablation)Compressive Transformer · Enwik8 testBPC: 1.01 -> 0.97Improved
increase compressed memory size from 1536 to 4096 (ablation)Compressive Transformer · Enwik8 testBPC: 0.97 -> 1.00Hurt
increase compressed memory size from 256 to 1536 (ablation)Compressive Transformer · WikiText-103 testperplexity: 18.2 -> 17.1Improved
reduce the frequency of optimizer updates near the end of trainingTransformerXL baseline · Enwik8BPC: 0.995 -> 0.984Improved

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