{"total":72,"rows":[{"id":"2412.06464","title":"Gated Delta Networks: Improving Mamba2 with Delta Rule","created":"2024-12-09","cited_by_count":532,"area":null},{"id":"2505.06708","title":"Gated Attention for Large Language Models: Non-linearity, Sparsity, and Attention-Sink-Free","created":"2025-05-10","cited_by_count":283,"area":"moe"},{"id":"2502.07529","title":"Training Deep Learning Models with Norm-Constrained LMOs","created":"2025-02-11","cited_by_count":192,"area":null},{"id":"2405.18392","title":"Scaling Laws and Compute-Optimal Training Beyond Fixed Training Durations","created":"2024-05-28","cited_by_count":170,"area":null},{"id":"2503.04482","title":"Generalized Interpolating Discrete Diffusion","created":"2025-03-06","cited_by_count":70,"area":"diffusion"},{"id":"2501.04697","title":"Grokking at the Edge of Numerical Stability","created":"2025-01-08","cited_by_count":46,"area":null},{"id":"2405.13698","title":"How to set AdamW's weight decay as you scale model and dataset size","created":"2024-05-22","cited_by_count":43,"area":null},{"id":"2505.13738","title":"Power Lines: Scaling Laws for Weight Decay and Batch Size in LLM Pre-training","created":"2025-05-19","cited_by_count":43,"area":null},{"id":"2502.15938","title":"Straight to Zero: Why Linearly Decaying the Learning Rate to Zero Works Best for LLMs","created":"2025-02-21","cited_by_count":35,"area":null},{"id":"2408.11804","title":"Approaching Deep Learning through the Spectral Dynamics of Weights","created":"2024-08-21","cited_by_count":31,"area":null},{"id":"2509.14786","title":"Pre-training under infinite compute","created":"2025-09-18","cited_by_count":29,"area":null},{"id":"2409.15156","title":"Rethinking Conventional Wisdom in Machine Learning: From Generalization to Scaling","created":"2024-09-23","cited_by_count":23,"area":"lr"},{"id":"2506.02285","title":"Why Gradients Rapidly Increase Near the End of Training","created":"2025-06-02","cited_by_count":22,"area":null},{"id":"2510.19093","title":"Weight Decay may matter more than muP for Learning Rate Transfer in Practice","created":"2025-10-21","cited_by_count":20,"area":null},{"id":"2506.01913","title":"Generalized Gradient Norm Clipping & Non-Euclidean $(L_0,L_1)$-Smoothness","created":"2025-06-02","cited_by_count":17,"area":null},{"id":"2606.16899","title":"Fantastic Pretraining Optimizers and Where to Find Them II: Hyperball Optimization","created":"2026-06-15","cited_by_count":16,"area":null},{"id":"2402.09816","title":"Mind the Modality Gap: Towards a Remote Sensing Vision-Language Model via Cross-modal Alignment","created":"2024-02-15","cited_by_count":15,"area":"mllm"},{"id":"2410.03249","title":"How Much Can We Forget about Data Contamination?","created":"2024-10-04","cited_by_count":14,"area":null},{"id":"2505.20172","title":"A Theoretical Framework for Grokking: Interpolation followed by Riemannian Norm Minimisation","created":"2025-05-26","cited_by_count":13,"area":null},{"id":"2506.14562","title":"AlphaDecay: Module-wise Weight Decay for Heavy-Tailed Balancing in LLMs","created":"2025-06-17","cited_by_count":13,"area":null},{"id":"2510.12402","title":"Cautious Weight Decay","created":"2025-10-14","cited_by_count":13,"area":null},{"id":"2402.09240","title":"Switch EMA: A Free Lunch for Better Flatness and Sharpness","created":"2024-02-14","cited_by_count":9,"area":null},{"id":"2506.13234","title":"The Butterfly Effect: Neural Network Training Trajectories Are Highly Sensitive to Initial Conditions","created":"2025-06-16","cited_by_count":9,"area":null},{"id":"2509.25087","title":"Scaling with Collapse: Efficient and Predictable Training of LLM Families","created":"2025-09-29","cited_by_count":9,"area":null},{"id":"2501.12243","title":"FOCUS: First Order Concentrated Updating Scheme","created":"2025-01-21","cited_by_count":8,"area":null},{"id":"2505.22785","title":"Navigating the Latent Space Dynamics of Neural Models","created":"2025-05-28","cited_by_count":8,"area":null},{"id":"2510.15262","title":"Robust Layerwise Scaling Rules by Proper Weight Decay Tuning","created":"2025-10-17","cited_by_count":8,"area":null},{"id":"2601.16979","title":"A Scalable Measure of Loss Landscape Curvature for Analyzing the Training Dynamics of LLMs","created":"2026-01-23","cited_by_count":8,"area":null},{"id":"2502.09598","title":"GAIA: A Global, Multi-modal, Multi-scale Vision-Language Dataset for Remote Sensing Image Analysis","created":"2025-02-13","cited_by_count":7,"area":"mllm"},{"id":"2504.17160","title":"OUI Need to Talk About Weight Decay: A New Perspective on Overfitting Detection","created":"2025-04-24","cited_by_count":7,"area":null},{"id":"2401.11600","title":"Understanding the Generalization Benefits of Late Learning Rate Decay","created":"2024-01-21","cited_by_count":6,"area":null},{"id":"2410.22113","title":"Where Do Large Learning Rates Lead Us?","created":"2024-10-29","cited_by_count":6,"area":null},{"id":"2408.09414","title":"Clustering and Alignment: Understanding the Training Dynamics in Modular Addition","created":"2024-08-18","cited_by_count":5,"area":null},{"id":"2311.14387","title":"Achieving Margin Maximization Exponentially Fast via Progressive Norm Rescaling","created":"2023-11-24","cited_by_count":4,"area":null},{"id":"2410.11985","title":"The Fair Language Model Paradox","created":"2024-10-15","cited_by_count":4,"area":null},{"id":"2410.14602","title":"Data Diversity as Implicit Regularization: How Does Diversity Shape the Weight Space of Deep Neural Networks?","created":"2024-10-18","cited_by_count":4,"area":null},{"id":"2505.23489","title":"SGD as Free Energy Minimization: A Thermodynamic View on Neural Network Training","created":"2025-05-29","cited_by_count":4,"area":null},{"id":"2509.25049","title":"Efficient Hyperparameter Tuning via Trajectory Invariance Principle","created":"2025-09-29","cited_by_count":4,"area":null},{"id":"2602.01642","title":"The Effect of Mini-Batch Noise on the Implicit Bias of Adam","created":"2026-02-02","cited_by_count":4,"area":null},{"id":"2602.11137","title":"Weight Decay Improves Language Model Plasticity","created":"2026-02-11","cited_by_count":4,"area":null},{"id":"2505.11434","title":"Controlling the Flow: Stability and Convergence for Stochastic Gradient Descent with Decaying Regularization","created":"2025-05-16","cited_by_count":3,"area":null},{"id":"2602.20433","title":"Disentangling Geometry, Performance, and Training in Language Models","created":"2026-02-24","cited_by_count":3,"area":null},{"id":"2606.25971","title":"Improving Neural Network Training by Decoupling the Magnitude and Direction of Weight Vectors","created":"2026-06-24","cited_by_count":3,"area":"moe"},{"id":"2608.24814","title":"Effective Learning Rate Governs Loss Dynamics in Language Model Pretraining","created":"2026-08-25","cited_by_count":3,"area":null},{"id":"2403.07379","title":"Hallmarks of Optimization Trajectories in Neural Networks: Directional Exploration and Redundancy","created":"2024-03-12","cited_by_count":2,"area":null},{"id":"2510.04686","title":"How does the optimizer implicitly bias the model merging loss landscape?","created":"2025-10-06","cited_by_count":2,"area":null},{"id":"2510.23912","title":"Key and Value Weights Are Probably All You Need: On the Necessity of the Query, Key, Value weight Triplet in Self-Attention Transformers","created":"2025-10-27","cited_by_count":2,"area":null},{"id":"2601.04890","title":"Learnable Multipliers: Freeing the Scale of Language Model Matrix Layers","created":"2026-01-08","cited_by_count":2,"area":null},{"id":"2601.19611","title":"Explicit Multi-head Attention for Inter-head Interaction in Large Language Models","created":"2026-01-27","cited_by_count":2,"area":"inference"},{"id":"2602.16740","title":"Quantifying LLM Attention-Head Stability: Implications for Circuit Universality","created":"2026-02-17","cited_by_count":2,"area":"mechinterp"},{"id":"2603.13381","title":"Beyond Linearity in Attention Projections: The Case for Nonlinear Queries","created":"2026-03-11","cited_by_count":2,"area":null},{"id":"2605.10878","title":"Neural Weight Norm = Kolmogorov Complexity","created":"2026-05-11","cited_by_count":2,"area":null},{"id":"2605.18528","title":"Tight Complexity of Muon under Heavy-Tailed Noise","created":"2026-05-18","cited_by_count":2,"area":null},{"id":"2605.20441","title":"Weight Decay Regimes in Grokking Transformers: Cheap Online Diagnostics","created":"2026-05-19","cited_by_count":2,"area":null},{"id":"2505.22578","title":"Favorability of Loss Landscape with Weight Decay Requires Both Large Overparametrization and Initialization","created":"2025-05-28","cited_by_count":1,"area":null},{"id":"2507.04683","title":"Recovering Plasticity of Neural Networks via Soft Weight Rescaling","created":"2025-07-07","cited_by_count":1,"area":null},{"id":"2512.08217","title":"Correction of Decoupled Weight Decay","created":"2025-12-09","cited_by_count":1,"area":null},{"id":"2605.04418","title":"Demystifying Manifold Constraints in LLM Pre-training","created":"2026-05-06","cited_by_count":1,"area":null},{"id":"2605.15290","title":"GQA-{\\mu}P: The maximal parameterization update for grouped query attention","created":"2026-05-14","cited_by_count":1,"area":null},{"id":"2605.16622","title":"Does Weight Decay Enhance Training Stability?","created":"2026-05-15","cited_by_count":1,"area":null},{"id":"2605.18845","title":"First-Passage Prediction of Grokking Delay: ACalibrated Law under AdamW with Causal Validation","created":"2026-05-13","cited_by_count":1,"area":null},{"id":"2510.08341","title":"Post-Norm can Resharpen Attention","created":"2025-10-09","cited_by_count":0,"area":null},{"id":"2511.13654","title":"Tuning for Two Adversaries: Enhancing the Robustness Against Transfer and Query-Based Attacks using Hyperparameter Tuning","created":"2025-11-17","cited_by_count":0,"area":null},{"id":"2601.16736","title":"A Step to Decouple Optimization in 3DGS","created":"2026-01-23","cited_by_count":0,"area":null},{"id":"2602.16213","title":"Graph neural network for colliding particles with an application to sea ice floe modeling","created":"2026-02-18","cited_by_count":0,"area":null},{"id":"2604.23434","title":"When Does Removing LayerNorm Help? Activation Bounding as a Regime-Dependent Implicit Regularizer","created":"2026-04-25","cited_by_count":0,"area":null},{"id":"2608.15483","title":"Measuring Structured Predictability in Neural Training Dynamics: A Cross-Regime Study","created":"2026-08-16","cited_by_count":0,"area":null},{"id":"2608.28948","title":"From the Loss Landscape to Diverse Feature Learning in Neural Networks","created":"2026-08-28","cited_by_count":0,"area":null},{"id":"2609.05275","title":"Don't Drop Dropout: Optimizing Layer Sparsity for Efficient LLM Training and Inference","created":"2026-09-04","cited_by_count":0,"area":"inference"},{"id":"2609.05730","title":"Bigger Text Encoders Can Hurt CLIP Zero-Shot Performance","created":"2026-09-04","cited_by_count":0,"area":null},{"id":"2609.07017","title":"HyperTransfer: Understanding the Equivalence between Base Optimizer and Hyperball","created":"2026-09-07","cited_by_count":0,"area":null},{"id":"2609.09116","title":"When Does Scale-Invariant Optimization Become Unstable? An Exact Schedule Law with Weight Decay","created":"2026-09-08","cited_by_count":0,"area":null}],"next_cursor":null}