AI Paper Atlas

Research map / Architectures and efficiency / Attention and transformers

Understanding How Encoder-Decoder Architectures Attend

Kyle Aitken, Vinay Venkatesh Ramasesh, Yuan Cao, Niru Maheswaranathan

NeurIPS 2021 · Architectures and efficiency › Attention and transformers · approach: Transformer expressivity and circuits · 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
add two input-dependent terms to the temporal-only alignment approximationalignment score approximation · eSCANalignment variance captured (%): >77% -> >87%Improved
subtract temporal component from hidden states (AED)encoder hidden states · eSCANword-state variance drop (%): 90%Improved
subtract temporal component from hidden states (AED)encoder hidden states · English-French translationword-state variance drop (%): 8%No clear effect
subtract temporal component from hidden states (AO)encoder hidden states · English-French translationword-state variance drop (%): 25%No clear effect
swap GRU cell for LSTM in AEDAED · one-to-one taskattention patternMixed
swap GRU for UGRNN or LSTM cellsAED/AO · one-to-one tasktest accuracy (%): 100%No 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