AI Paper Atlas

Research map / Architectures and efficiency / RNNs and normalization

Two-Stage Unit Tying for Simplifying Differentiable Logic Gate Networks

Seungheon Lee, Jeongmin Sun, Jaeyong Chung

ICML 2026 · Architectures and efficiency › RNNs and normalization · approach: Differentiable optimization layers

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 Stage-B Binary Split refinement to Stage-A screening (ablation)Ours · CIFAR-10(M) 50% tiedaccuracy (%): 59.22 -> 61.44Improved
add Stage-B Binary Split refinement to Stage-A screening (ablation)Ours · MNIST(M) 50% tiedaccuracy (%): 91.71 -> 96.20Improved
add post-tying fine-tuning (ablation)Ours · CIFAR-10(M) 50% tiedaccuracy (%): 61.44 -> 69.22Improved
use Binary Split refinement instead of Tail Rescore (ablation)Ours · CIFAR-10(M) 30% tiedaccuracy (%): 66.33 -> 67.30Improved
use Binary Split refinement instead of Full-set Rescore (ablation)Ours · CIFAR-10(M) 30% tiedaccuracy (%): 66.88 -> 67.30Improved
use logit-level MSE instead of KL with tau=1 (ablation)Ours · CIFAR-10(M) 50% tied, Stage-Aaccuracy (%): 55.88 -> 59.22Improved
raise KL temperature from 1 to 20 (ablation)Ours · CIFAR-10(M) 50% tied, Stage-Aaccuracy (%): 55.88 -> 59.14Improved
Gauss-Newton screening plus refinement vs FD-DL (N=128)Ours · CIFAR-10(M) 50% tiedaccuracy (%): 56.05 -> 61.44Improved
Ours vs Fisher-TL saliencyOurs · CIFAR-10(M) 50% tiedaccuracy (%): 53.97 -> 61.44Improved
Ours vs ConstProb saliencyOurs · MNIST(M) 50% tiedaccuracy (%): 31.18 -> 96.20Improved
increase FD-DL calibration size N from 16 to 64 (ablation)FD-DL · CIFAR-10(M) 30% tiedaccuracy (%): 66.38 -> 67.30Improved
use Stage-A only vs FD-DL at 30% tiedOurs (Stage-A only) · CIFAR-10(M) 30% tiedaccuracy (%): 66.49 -> 67.19Mixed

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