AI Paper Atlas

Research map / Architectures and efficiency / Attention and transformers

Push, Pop, Parallelize: Stack-Augmented Linear Attention via the Delta Rule

Anh T Nguyen, Saleh Momeni, Ashutosh Chaubey, Changnan Xiao, Bing Liu

ICML 2026 · Architectures and efficiency › Attention and transformers · approach: Linear attention approximations

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 a differentiable stack to DeltaNetDeltaNet · Formal languages (average over 8 tasks)accuracy: 65.5 -> 79.4Improved
add a differentiable stack to DeltaNetDeltaNet · Formal languages: Stack Manipulationaccuracy: 56.6 -> 98.8Improved
add a differentiable stack to DeltaNetDeltaNet · Formal languages: Parityaccuracy: 64.3 -> 96.8Improved
add a differentiable stack to DeltaNetDeltaNet · Formal languages: Dyck Mod Arith. w/ Bracketsaccuracy: 33.2 -> 35.2No clear effect
add a differentiable stack to DeltaNetDeltaNet · Formal languages: Cycle Navigationaccuracy: 46.7 -> 46.4No clear effect
add a differentiable stack to DeltaNetDeltaNet · MAD benchmark (average)average score: 71.8 -> 72.2No clear effect
add a differentiable stack to DeltaNet (s = 8, s = 16)DeltaNet · FineWeb-Edu LM, 340M params / 15B tokensperplexity and zero-shot average: 45.5 -> 45.9 (s=8), 46.2 (s=16) avg accNo clear effect
add a differentiable stack to DeltaNet (s = 8)DeltaNet · FineWeb-Edu LM, 760M params / 30B tokensaverage zero-shot accuracy: 51.4 -> 52.0Improved
add a differentiable stack to DeltaNetDeltaNet 340M · CodeParrot and OpenThoughts-114k-Math length extrapolation to 16kper-token lossImproved
increase stack size from 8 to 48 (ablation)DeltaStack · Formal languages: Parity, Reverse String, Stack ManipulationaccuracyImproved
increase stack size (ablation)DeltaStack · Formal languages: state tracking (Parity)accuracyNo clear effect
add a differentiable stack to DeltaNetDeltaNet 1.3B · Peak training memory, A100peak memory (GiB): 41.3 -> 45.1Hurt

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