The Weight Gram Matrix Captures Sequential Feature Linearization in Deep Networks
cs.LG, cs.AI
Submitted: 2026-05-07
Updated: 2026-11-02
Code: https://github.com/cth127/GramLin
Terminology
Sources
- Understanding intermediate layers using linear classifier probes
- The Computational Advantage of Depth: Learning High-Dimensional Hierarchical Functions with Gradient Descent
- Gaussian Error Linear Units (GELUs)
- Adam: A Method for Stochastic Optimization
- Auto-Encoding Variational Bayes
- Early Directional Convergence in Deep Homogeneous Neural Networks for Small Initializations
- Muon is Scalable for LLM Training
- Efficient Estimation of Word Representations in Vector Space
- Neural Collapse under Gradient Flow on Shallow ReLU Networks for Orthogonally Separable Data
- Grokking: Generalization Beyond Overfitting on Small Algorithmic Datasets
- Representation Engineering: A Top-Down Approach to AI Transparency
Related papers
- FlexMoE: Learning One-for-All Multi-Budget Expert Pruning for Flexible MoE Deployment
- OptiMind: Improving Small Language Models for Optimization Modeling through Formulation Guidance and Multi-Turn Learning
- Staying in the Sweet Spot: Boost Reasoning Evolution via Responsive Rollout Difficulty Manipulation
- Examining the Impact of Data and Training Methods on Emergent and Subliminal Misalignment
- Rethinking Black-Box On-Policy Distillation: A Rubric-based Analysis
- (Mis)generalization of Helpful-Only Fine-tuning