AI Paper Atlas

Research map / Architectures and efficiency / RNNs and normalization

KLAS: Using Similarity to Stitch Neural Networks for an Improved Accuracy-Efficiency Tradeoff

Debopam Sanyal, Anantharaman S. Iyer, Alind Khare, Trisha Jain, Akshay Jajoo, Myungjin Lee, James Clayton Kerce, Alexey Tumanov

ICLR 2026 · Architectures and efficiency › RNNs and normalization · approach: Dataset condensation and flow estimation

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
use KL-divergence-based stitch selection (KLAS) instead of heuristic selectionSN-Net · ImageNet-1K, SwinAUC: 0.8345 -> 0.8950Improved
select stitches with KLAS on ImageNet-1KSN-Net · ImageNet-1Ktop-1 accuracy / FLOPs: up to 1.21% higher accuracy, or 1.33x fewer FLOPsImproved
use far stitching (Swin Ti-B) instead of nearest stitching (Ti-S-B) (ablation)Swin stitching · ImageNet-1Kaccuracy at equal FLOPs: up to 0.9%Improved
use far stitching (Ti-B) instead of nearest stitching for DeiT (ablation)DeiT stitching · ImageNet-1Kaccuracy-efficiency curveHurt
use KLAS block selection with SN-Net's anchorsSN-Net · ImageNet-1K, Swin and DeiTAUCImproved
use naive Min-KL selection instead of target-aware stitch score (ablation)KLAS · DeiT and Swinaccuracy-efficiency curveHurt
use KLAS on dense predictionSN-Net · ADE20K, Mask2Former Swin-T/BmIoU: up to +0.9% mIoUImproved
use KLAS instead of ESTA layer-count heuristic for LLM stitchingESTA · TruthfulQA, Llama 3.2 1B to 3BROUGE-1 and ROUGE-2: up to 0.017 ROUGE-1 and 0.033 ROUGE-2Improved
use KLAS stitched models instead of early-exit cascadesSwin-S and Swin-B cascade · ImageNet-1Ktop-1 accuracy: over 0.4% above best cascade at more than 12 GFLOPsImproved
use KLAS instead of cascade for ResNet-18 to Swin-Bcascade · ImageNet-1Ktop-1 accuracy: 76.13 -> 79.60 at config 1Improved
train probes with shuffled labels (ablation)stitch score · SwinPearson correlation with accuracy drop: 0.84 -> 0.19Hurt
use threshold tau of 10% instead of 5% (ablation)KLAS · Swin, 20 bucketsaccuracy and AUCHurt

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