AI Paper Atlas

Research map / Architectures and efficiency / Attention and transformers

Treeformer: Dense Gradient Trees for Efficient Attention Computation

Lovish Madaan, Srinadh Bhojanapalli, Himanshu Jain, Prateek Jain

ICLR 2023 · Architectures and efficiency › Attention and transformers · approach: Linear attention approximations · 0 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 full attention with coarse tree attentionTreeformer on LRA · LRA (4-task average)attention FLOPs: 100% -> 3.05%Improved
Replace full attention with coarse tree attentionTreeformer on LRA · LRA (4-task average)score: 50.14 -> 50.09No clear effect
Replace full attention with fine-grained tree attentionTreeformer on LRA · LRA (4-task average)score: 50.14 -> 50.31Improved
Use k-ary instead of binary trees (ablation)Treeformer TF-A · BERT-Base fine-tuningGLUE average: 80.05 -> 81.25Improved
Use k-ary instead of binary trees (ablation)Treeformer TF-A · PubMed summarizationROUGE-1: 45.02 -> 45.35Mixed
Increase TC-A layers from 6 to 9 (ablation)Treeformer TC-A · BERT-Base fine-tuningGLUE average: 80.55 -> 79.51Hurt
Increase TC-A layers from 6 to 12 (ablation)Treeformer TC-A · BERT-Base fine-tuningGLUE average: 80.55 -> 71.39Hurt
Use bootstrapped training rather than training the full tree model from scratch (ablation)Treeformer TF-A · BERT pre-trainingMLM accuracyImproved
Extend TC-A bootstrapping to 12 decision-tree layers rather than stopping at 6 (ablation)Treeformer TC-A · BERT pre-trainingMLM accuracyHurt
Use TC-A rather than TF-A (ablation)Treeformer on PubMed summarization · PubMed summarizationROUGE-1: 45.02 -> 40.23Hurt

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 0 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