·
DataBubble
  • Home
  • Models
  • News
  • Compare
  • Boards
  • Pricing
  • About
  • Newsletter
  • Methodology
  • Contact
Latest
SFMambaNet: Spectral-Frequency Enhanced Selective State Space Model for Correspondence Pruning1h◆MSTN: A Lightweight and Fast Model for General TimeSeries Analysis1h◆Optical-Guided Neural Collapse for SAR Few-Shot Class Incremental Learning1h◆Dynamic Infilling Anchors for Format-Constrained Generation in Diffusion Large Language Models1h◆Temporal Order Matters for Agentic Memory: Segment Trees for Long-Horizon Agents1h◆GIPO: Gaussian Importance Sampling Policy Optimization1h◆Specialization of softmax attention heads: insights from the high-dimensional single-location model1h◆On the Robustness of Langevin Dynamics to Score Function Error1h◆Level Up: Defining and Exploiting Transitional Problems for Curriculum Learning1h◆LeWorldModel: Stable End-to-End Joint-Embedding Predictive Architecture from Pixels1h◆From Causal Discovery to Dynamic Causal Inference in Neural Time Series1h◆Why Muon Outperforms Adam: A Curvature Perspective1h◆Vision Hopfield Memory Networks1h◆Provably Auditable and Safe LLM Agents from Human-Authored Ontologies1h◆Algebraic Diversity: Group-Theoretic Spectral Estimation from Single Observations1h◆Stochastic-Dimension Frozen Sampled Neural Network for High-Dimensional Gross-Pitaevskii Equations on Unbounded Domains1h◆SOLARIS: Speculative Offloading of Latent-bAsed Representation for Inference Scaling1h◆The Topological Trouble With Transformers1h◆Adaptive Head Budgeting for Efficient Multi-Head Attention1h◆Evaluating CUDA Tile for AI Workloads on Hopper and Blackwell GPUs1h◆SFMambaNet: Spectral-Frequency Enhanced Selective State Space Model for Correspondence Pruning1h◆MSTN: A Lightweight and Fast Model for General TimeSeries Analysis1h◆Optical-Guided Neural Collapse for SAR Few-Shot Class Incremental Learning1h◆Dynamic Infilling Anchors for Format-Constrained Generation in Diffusion Large Language Models1h◆Temporal Order Matters for Agentic Memory: Segment Trees for Long-Horizon Agents1h◆GIPO: Gaussian Importance Sampling Policy Optimization1h◆Specialization of softmax attention heads: insights from the high-dimensional single-location model1h◆On the Robustness of Langevin Dynamics to Score Function Error1h◆Level Up: Defining and Exploiting Transitional Problems for Curriculum Learning1h◆LeWorldModel: Stable End-to-End Joint-Embedding Predictive Architecture from Pixels1h◆From Causal Discovery to Dynamic Causal Inference in Neural Time Series1h◆Why Muon Outperforms Adam: A Curvature Perspective1h◆Vision Hopfield Memory Networks1h◆Provably Auditable and Safe LLM Agents from Human-Authored Ontologies1h◆Algebraic Diversity: Group-Theoretic Spectral Estimation from Single Observations1h◆Stochastic-Dimension Frozen Sampled Neural Network for High-Dimensional Gross-Pitaevskii Equations on Unbounded Domains1h◆SOLARIS: Speculative Offloading of Latent-bAsed Representation for Inference Scaling1h◆The Topological Trouble With Transformers1h◆Adaptive Head Budgeting for Efficient Multi-Head Attention1h◆Evaluating CUDA Tile for AI Workloads on Hopper and Blackwell GPUs1h◆
News/StableGrad: Backward Scale Control without Batch Normalization
arxiv
PublishedMay 21, 2026 at 4:00 AM

StableGrad: Backward Scale Control without Batch Normalization

Source
arxiv.orgfull article ↗
Read on arxiv→
Publisher summary· verbatim

arXiv:2605.19856v1 Announce Type: cross Abstract: Training very deep neural networks requires controlling the propagation of magnitudes across depth. Without such control, activations and gradients may vanish, explode, or enter unstable regimes that make optimization fail. Modern architectures often

Stay posted· Newsletter

A 5-min weekly brief — top movers, price watch, story of the week.

// no spam · unsubscribe one-click · free forever

Discussion
Source
↗
arxiv
Read original ↗All from arxiv →

No replies yet. Be first.

Source
↗
arxiv
Read original ↗All from arxiv →

Related coverage

More from ARXIV
arxivSFMambaNet: Spectral-Frequency Enhanced Selective State Space Model for Correspondence Pruning1harxivMSTN: A Lightweight and Fast Model for General TimeSeries Analysis1harxivOptical-Guided Neural Collapse for SAR Few-Shot Class Incremental Learning1harxivDynamic Infilling Anchors for Format-Constrained Generation in Diffusion Large Language Models1h
The Bubble Brief
WEEKLY

Read AI insights every Tuesday — top movers, new releases, story of the week.

// no spam · unsubscribe one-click · free forever

Originally published on arxiv ↗
HomeModelsNews