AI Paper Atlas

Research map / Architectures and efficiency / Pruning and efficient tuning

CR-Net: Scaling Parameter-Efficient Training with Cross-Layer Low-Rank Structure

Boao Kong, Junzhu Liang, Yuxi Liu, Renjia Deng, Kun Yuan

ICLR 2026 · Architectures and efficiency › Pruning and efficient tuning · approach: Low-rank compression and merging · 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
use a full-rank first layer instead of low-rank (ablation)CR-Net · LLaMA-2 350M C4validation perplexity: 19.68 -> 18.95Improved
make the scaling factor beta learnable instead of fixed (ablation)CR-Net · LLaMA-2 350M C4training stabilityImproved
use higher ranks in middle layers and lower in side layers (ablation)CR-Net · LLaMA-2 350M C4validation perplexityImproved
use previous-layer residual instead of learnable first-layer skip (Learnable-ResFormer) or DenseFormer (ablation)CR-Net · LLaMA-2 1B C4validation perplexityImproved
replace full-rank training with CR-Net at aligned parametersCR-Net · LLaMA-2 1B C4validation perplexity: 15.56 -> 15.22Improved
CR-Net vs CoLA at aligned parametersCR-Net · LLaMA-2 60M C4validation perplexity: 34.04 -> 32.76Improved
CR-Net vs full-rank at 350MCR-Net · LLaMA-2 350M C4validation perplexity: 18.80 -> 18.95Hurt
CR-Net with re-computation vs CoLA-MCR-Net · LLaMA-2 7B C4 80K stepsvalidation perplexity: 13.82 -> 13.72Improved
CR-Net with re-computation vs 8-bit AdamCR-Net · LLaMA-2 13B C4 40K stepsvalidation perplexity: 17.85 -> 18.12Hurt
use tailored re-computation instead of vanilla gradient checkpointingCR-Net · LLaMA2-7B activation memorymemory (GB): 51.22 -> 23.35Improved
CR-Net re-computation vs CoLA-M activation memoryCR-Net · LLaMA2-7B activation memorymemory (GB): 24.78 -> 23.35Improved
CR-Net at aligned memory vs ApolloCR-Net · LLaMA-2 1B C4validation perplexity: 14.20 -> 14.05Improved

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