Research map / Architectures and efficiency / RNNs and normalization
Going Beyond Linear Mode Connectivity: The Layerwise Linear Feature Connectivity
Zhanpeng Zhou, Yongyi Yang, Xiaojiang Yang, Junchi Yan, Wei Hu
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 changed | Where | Measured | Effect |
|---|---|---|---|
| use spawning to obtain LMC-connected modes instead of independently trained modes | neural networks trained with spawning · MNIST, CIFAR-10, and Tiny-ImageNet; reported architectures | 1 - cosine similarity between interpolated-model features and interpolated endpo | Improved |
| use permutation matching to obtain LMC-connected modes instead of leaving independently trained modes unaligned | neural networks trained with permutation matching · CIFAR-10 and other reported image-classification settings | 1 - cosine similarity between interpolated-model features and interpolated endpo | Improved |
| use spawning to obtain linearly connected modes instead of independently trained modes | ReLU networks trained with spawning · Reported datasets and architectures | weak-additivity distance | Improved |
| apply activation matching or weight matching before comparing independently trained modes | Git Re-Basin permutation method · Reported datasets and architectures | weak-additivity distance | Improved |
| use spawning or permutation to obtain LLFC modes instead of comparing independently trained modes | neural networks trained with spawning or permutation · MNIST and CIFAR-10; reported architectures | commutativity distance | Improved |
Closest papers by meaning
- Linear Mode Connectivity in Differentiable Tree Ensembles (ICLR 2025 · similarity 0.87)
- On Linear Mode Connectivity of Mixture-of-Experts Architectures (NeurIPS 2025 · similarity 0.87)
- Layer-wise linear mode connectivity (ICLR 2024 · similarity 0.87)
- Understanding Mode Connectivity via Parameter Space Symmetry (ICML 2025 · similarity 0.87)
- Generalized Linear Mode Connectivity for Transformers (NeurIPS 2025 · similarity 0.86)
- Three Mechanisms of Feature Learning in a Linear Network (ICLR 2025 · similarity 0.86)
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.
