Image de couverture de livre

Spectral Edge Dynamics of Training Trajectories: Signal--Noise Geometry Across Scales

Fuente: arXiv

Enregistré dans:

Détails bibliographiques
Auteur principal:	Xu, Yongzhong
Format:	Preprint
Publié:	2026
Sujets:	Machine Learning Artificial Intelligence
Accès en ligne:	Acceder al recurso
Tags:	Ajouter un tag Pas de tags, Soyez le premier à ajouter un tag!

Documents similaires

Spectral Edge Dynamics: An Analytical-Empirical Study of Phase Transitions in Neural Network Training
par: Xu, Yongzhong
Publié: (2026)

Spectral Edge Dynamics Reveal Functional Modes of Learning
par: Xu, Yongzhong
Publié: (2026)

Early-Warning Signals of Grokking via Loss-Landscape Geometry
par: Xu, Yongzhong
Publié: (2026)

When Do Attention Circuits Form? Developmental Trajectories of Capability and Attention-Sink Emergence Across Three 1B-ClassArchitectures
par: Xu, Yongzhong
Publié: (2026)

Optimizer-Induced Low-Dimensional Drift and Transverse Dynamics in Transformer Training
par: Xu, Yongzhong
Publié: (2026)

Gradient-Direction Sensitivity Reveals Linear-Centroid Coupling Hidden by Optimizer Trajectories
par: Xu, Yongzhong
Publié: (2026)

The Geometry of Multi-Task Grokking: Transverse Instability, Superposition, and Weight Decay Phase Structure
par: Xu, Yongzhong
Publié: (2026)

Feature Repulsion and Spectral Lock-in: An Empirical Study of Two-Layer Network Grokking
par: Xu, Yongzhong
Publié: (2026)

Spectral Probe-Circuits: A Three-Step Recipe for Identifying Attention-Head Circuits in Pretrained Transformers
par: Xu, Yongzhong
Publié: (2026)

Low-Dimensional and Transversely Curved Optimization Dynamics in Grokking
par: Xu, Yongzhong
Publié: (2026)

Global Low-Rank, Local Full-Rank: The Holographic Encoding of Learned Algorithms
par: Xu, Yongzhong
Publié: (2026)

Low-Dimensional Execution Manifolds in Transformer Learning Dynamics: Evidence from Modular Arithmetic Tasks
par: Xu, Yongzhong
Publié: (2026)

From Dispersion to Attraction: Spectral Dynamics of Hallucination Across Whisper Model Scales
par: Viakhirev, Ivan, et autres
Publié: (2026)

The Lifecycle of the Spectral Edge: From Gradient Learning to Weight-Decay Compression
par: Xu, Yongzhong
Publié: (2026)

AOI: Turning Failed Trajectories into Training Signals for Autonomous Cloud Diagnosis
par: Yang, Pei, et autres
Publié: (2026)

LLM Reasoning as Trajectories: Step-Specific Representation Geometry and Correctness Signals
par: Sun, Lihao, et autres
Publié: (2026)

Dynamics of the Transformer Residual Stream: Coupling Spectral Geometry to Network Topology
par: Fernando, Jesseba, et autres
Publié: (2026)

Dynamic Spectral Backpropagation for Efficient Neural Network Training
par: Muthuraman, Mannmohan
Publié: (2025)

Introspective X Training: Feedback Conditioning Improves Scaling Across all LLM Training Stages
par: Cui, Brandon, et autres
Publié: (2026)

Adaptive Batch Sizes Using Non-Euclidean Gradient Noise Scales for Stochastic Sign and Spectral Descent
par: Naganuma, Hiroki, et autres
Publié: (2026)

Informed Spectral Normalized Gaussian Processes for Trajectory Prediction
par: Schlauch, Christian, et autres
Publié: (2024)

Revealing Modular Gradient Noise Imbalance in LLMs: Calibrating Adam via Signal-to-Noise Ratio
par: Wen, Ziqing, et autres
Publié: (2026)

Do Post-Training Algorithms Actually Differ? A Controlled Study Across Model Scales Uncovers Scale-Dependent Ranking Inversions
par: Li, Xiaoyi
Publié: (2026)

Clarifying Shampoo: Adapting Spectral Descent to Stochasticity and the Parameter Trajectory
par: Eschenhagen, Runa, et autres
Publié: (2026)

Controlled LLM Training on Spectral Sphere
par: Xie, Tian, et autres
Publié: (2026)

The Malignant Tail: Spectral Segregation of Label Noise in Over-Parameterized Networks
par: Wang, Zice
Publié: (2026)

The Geometries of Truth Are Orthogonal Across Tasks
par: Azizian, Waiss, et autres
Publié: (2025)

Spectral Theory for Edge Pruning in Asynchronous Recurrent Graph Neural Networks
par: Bessone, Nicolas
Publié: (2025)

Enhancing Adversarial Training via Reweighting Optimization Trajectory
par: Huang, Tianjin, et autres
Publié: (2023)

Joint Training Across Multiple Activation Sparsity Regimes
par: Wang, Haotian
Publié: (2026)

Memory-Efficient LLM Training with Dynamic Sparsity: From Stability to Practical Scaling
par: Xiao, Qiao, et autres
Publié: (2026)

Intrinsic Training Signals for Federated Learning Aggregation
par: Fiorini, Cosimo, et autres
Publié: (2025)

Spurious Rewards: Rethinking Training Signals in RLVR
par: Shao, Rulin, et autres
Publié: (2025)

Scale-Agnostic Kolmogorov-Arnold Geometry in Neural Networks
par: Vanherreweghe, Mathew, et autres
Publié: (2025)

Ensemble of Pre-Trained Models for Long-Tailed Trajectory Prediction
par: Thuremella, Divya, et autres
Publié: (2025)

Spectral Superposition: A Theory of Feature Geometry
par: Ivanov, Georgi, et autres
Publié: (2026)

Spectral Geometry for Deep Learning: Compression and Hallucination Detection via Random Matrix Theory
par: Ettori, Davide
Publié: (2026)

On the Spectral Geometry of Cross-Modal Representations: A Functional Map Diagnostic for Multimodal Alignment
par: Sarkar, Krisanu
Publié: (2026)

IntTrajSim: Trajectory Prediction for Simulating Multi-Vehicle driving at Signalized Intersections
par: Ranjan, Yash, et autres
Publié: (2025)

Handling Label Noise via Instance-Level Difficulty Modeling and Dynamic Optimization
par: Zhang, Kuan, et autres
Publié: (2025)