{"total":427,"rows":[{"id":"2402.12354","title":"LoRA+: Efficient Low Rank Adaptation of Large Models","created":"2024-02-19","cited_by_count":377,"area":null},{"id":"2309.14322","title":"Small-scale proxies for large-scale Transformer training instabilities","created":"2023-09-25","cited_by_count":183,"area":null},{"id":"2303.01486","title":"Understanding plasticity in neural networks","created":"2023-03-02","cited_by_count":174,"area":null},{"id":"2303.06296","title":"Stabilizing Transformer Training by Preventing Attention Entropy Collapse","created":"2023-03-11","cited_by_count":161,"area":null},{"id":"2209.15594","title":"Self-Stabilization: The Implicit Bias of Gradient Descent at the Edge of Stability","created":"2022-09-30","cited_by_count":155,"area":null},{"id":"2205.01445","title":"High-dimensional Asymptotics of Feature Learning: How One Gradient Step Improves the Representation","created":"2022-05-03","cited_by_count":150,"area":null},{"id":"2302.11055","title":"SGD learning on neural networks: leap complexity and saddle-to-saddle dynamics","created":"2023-02-21","cited_by_count":143,"area":null},{"id":"2106.06530","title":"Label Noise SGD Provably Prefers Flat Global Minimizers","created":"2021-06-11","cited_by_count":121,"area":null},{"id":"2304.13960","title":"Noise Is Not the Main Factor Behind the Gap Between SGD and Adam on Transformers, but Sign Descent Might Be","created":"2023-04-27","cited_by_count":114,"area":null},{"id":"2310.06110","title":"Grokking as the Transition from Lazy to Rich Training Dynamics","created":"2023-10-09","cited_by_count":108,"area":null},{"id":"2207.14484","title":"Adaptive Gradient Methods at the Edge of Stability","created":"2022-07-29","cited_by_count":101,"area":null},{"id":"2411.04330","title":"Scaling Laws for Precision","created":"2024-11-07","cited_by_count":95,"area":null},{"id":"2206.04817","title":"The Slingshot Mechanism: An Empirical Study of Adaptive Optimizers and the Grokking Phenomenon","created":"2022-06-10","cited_by_count":94,"area":null},{"id":"2302.07011","title":"A Modern Look at the Relationship between Sharpness and Generalization","created":"2023-02-14","cited_by_count":92,"area":null},{"id":"2204.01050","title":"Understanding the unstable convergence of gradient descent","created":"2022-04-03","cited_by_count":86,"area":null},{"id":"2206.07085","title":"Understanding the Generalization Benefit of Normalization Layers: Sharpness Reduction","created":"2022-06-14","cited_by_count":86,"area":null},{"id":"2208.12591","title":"On the Implicit Bias in Deep-Learning Algorithms","created":"2022-08-26","cited_by_count":86,"area":null},{"id":"2310.04415","title":"Why Do We Need Weight Decay in Modern Deep Learning?","created":"2023-10-06","cited_by_count":72,"area":null},{"id":"2206.04030","title":"High-dimensional limit theorems for SGD: Effective dynamics and critical scaling","created":"2022-06-08","cited_by_count":70,"area":null},{"id":"2210.05337","title":"SGD with Large Step Sizes Learns Sparse Features","created":"2022-10-11","cited_by_count":68,"area":null},{"id":"2405.14813","title":"Scalable Optimization in the Modular Norm","created":"2024-05-23","cited_by_count":68,"area":null},{"id":"2207.02628","title":"The alignment property of SGD noise and how it helps select flat minima: A stability analysis","created":"2022-07-06","cited_by_count":64,"area":null},{"id":"2106.04690","title":"Handcrafted Backdoors in Deep Neural Networks","created":"2021-06-08","cited_by_count":63,"area":null},{"id":"2109.14119","title":"Stochastic Training is Not Necessary for Generalization","created":"2021-09-29","cited_by_count":63,"area":null},{"id":"2207.12678","title":"Analyzing Sharpness along GD Trajectory: Progressive Sharpening and Edge of Stability","created":"2022-07-26","cited_by_count":63,"area":null},{"id":"2407.01800","title":"Normalization and effective learning rates in reinforcement learning","created":"2024-07-01","cited_by_count":58,"area":"deeprl"},{"id":"2206.10654","title":"On the Maximum Hessian Eigenvalue and Generalization","created":"2022-06-21","cited_by_count":57,"area":null},{"id":"2303.03108","title":"Gradient Norm Aware Minimization Seeks First-Order Flatness and Improves Generalization","created":"2023-03-03","cited_by_count":57,"area":null},{"id":"2210.03294","title":"Understanding Edge-of-Stability Training Dynamics with a Minimalist Example","created":"2022-10-07","cited_by_count":56,"area":null},{"id":"2409.17858","title":"How Feature Learning Can Improve Neural Scaling Laws","created":"2024-09-26","cited_by_count":56,"area":null},{"id":"2406.08447","title":"The Impact of Initialization on LoRA Finetuning Dynamics","created":"2024-06-12","cited_by_count":54,"area":null},{"id":"2306.00204","title":"Toward Understanding Why Adam Converges Faster Than SGD for Transformers","created":"2023-05-31","cited_by_count":53,"area":null},{"id":"2210.01513","title":"The Dynamics of Sharpness-Aware Minimization: Bouncing Across Ravines and Drifting Towards Wide Minima","created":"2022-10-04","cited_by_count":51,"area":null},{"id":"2212.07469","title":"Learning threshold neurons via the \"edge of stability\"","created":"2022-12-14","cited_by_count":51,"area":null},{"id":"2105.02716","title":"Noether's Learning Dynamics: Role of Symmetry Breaking in Neural Networks","created":"2021-05-06","cited_by_count":50,"area":null},{"id":"2406.09405","title":"Why Warmup the Learning Rate? Underlying Mechanisms and Improvements","created":"2024-06-13","cited_by_count":46,"area":null},{"id":"2105.06072","title":"Leveraging Non-uniformity in First-order Non-convex Optimization","created":"2021-05-13","cited_by_count":44,"area":null},{"id":"2204.11326","title":"Beyond the Quadratic Approximation: the Multiscale Structure of Neural Network Loss Landscapes","created":"2022-04-24","cited_by_count":43,"area":null},{"id":"2304.03408","title":"Dynamics of Finite Width Kernel and Prediction Fluctuations in Mean Field Neural Networks","created":"2023-04-06","cited_by_count":43,"area":null},{"id":"2305.11788","title":"Implicit Bias of Gradient Descent for Logistic Regression at the Edge of Stability","created":"2023-05-19","cited_by_count":42,"area":null},{"id":"2501.16371","title":"Optimizing the Optimizer for Physics-Informed Neural Networks and Kolmogorov-Arnold Networks","created":"2025-01-22","cited_by_count":42,"area":null},{"id":"2305.18411","title":"Feature-Learning Networks Are Consistent Across Widths At Realistic Scales","created":"2023-05-28","cited_by_count":41,"area":null},{"id":"2110.04369","title":"A Loss Curvature Perspective on Training Instability in Deep Learning","created":"2021-10-08","cited_by_count":40,"area":null},{"id":"2210.04860","title":"Second-order regression models exhibit progressive sharpening to the edge of stability","created":"2022-10-10","cited_by_count":40,"area":null},{"id":"2305.16292","title":"Sharpness-Aware Minimization Leads to Low-Rank Features","created":"2023-05-25","cited_by_count":38,"area":null},{"id":"2505.12082","title":"Model Merging in Pre-training of Large Language Models","created":"2025-05-17","cited_by_count":37,"area":"moe"},{"id":"2511.00674","title":"Isotropic Curvature Model for Understanding Deep Learning Optimization: Is Gradient Orthogonalization Optimal?","created":"2025-11-01","cited_by_count":37,"area":null},{"id":"2302.08982","title":"(S)GD over Diagonal Linear Networks: Implicit Regularisation, Large Stepsizes and Edge of Stability","created":"2023-02-17","cited_by_count":36,"area":null},{"id":"2405.16002","title":"Does SGD really happen in tiny subspaces?","created":"2024-05-25","cited_by_count":36,"area":null},{"id":"2305.16284","title":"DoWG Unleashed: An Efficient Universal Parameter-Free Gradient Descent Method","created":"2023-05-25","cited_by_count":35,"area":null},{"id":"2306.13253","title":"Predicting Grokking Long Before it Happens: A look into the loss landscape of models which grok","created":"2023-06-23","cited_by_count":35,"area":null},{"id":"2107.06608","title":"Continuous vs. Discrete Optimization of Deep Neural Networks","created":"2021-07-14","cited_by_count":34,"area":null},{"id":"2306.04815","title":"Catapults in SGD: spikes in the training loss and their impact on generalization through feature learning","created":"2023-06-07","cited_by_count":31,"area":null},{"id":"2509.15816","title":"On the Convergence of Muon and Beyond","created":"2025-09-19","cited_by_count":31,"area":null},{"id":"2209.11208","title":"A Closer Look at Learned Optimization: Stability, Robustness, and Inductive Biases","created":"2022-09-22","cited_by_count":30,"area":null},{"id":"2502.19002","title":"The Sharpness Disparity Principle in Transformers for Accelerating Language Model Pre-Training","created":"2025-02-26","cited_by_count":30,"area":null},{"id":"2202.00980","title":"Robust Training of Neural Networks Using Scale Invariant Architectures","created":"2022-02-02","cited_by_count":29,"area":null},{"id":"2604.21691","title":"There Will Be a Scientific Theory of Deep Learning","created":"2026-04-23","cited_by_count":29,"area":null},{"id":"2206.00730","title":"The Phenomenon of Policy Churn","created":"2022-06-01","cited_by_count":28,"area":null},{"id":"2210.15435","title":"Grokking phase transitions in learning local rules with gradient descent","created":"2022-10-26","cited_by_count":28,"area":null},{"id":"2501.06842","title":"SPAM: Spike-Aware Adam with Momentum Reset for Stable LLM Training","created":"2025-01-12","cited_by_count":28,"area":null},{"id":"2307.11108","title":"Flatness-Aware Minimization for Domain Generalization","created":"2023-07-20","cited_by_count":27,"area":null},{"id":"2309.00079","title":"On the Implicit Bias of Adam","created":"2023-08-31","cited_by_count":27,"area":null},{"id":"2305.15287","title":"The Crucial Role of Normalization in Sharpness-Aware Minimization","created":"2023-05-24","cited_by_count":26,"area":null},{"id":"2307.04204","title":"Trajectory Alignment: Understanding the Edge of Stability Phenomenon via Bifurcation Theory","created":"2023-07-09","cited_by_count":26,"area":null},{"id":"2312.00246","title":"Directions of Curvature as an Explanation for Loss of Plasticity","created":"2023-11-30","cited_by_count":26,"area":null},{"id":"2302.00849","title":"Implicit regularization in Heavy-ball momentum accelerated stochastic gradient descent","created":"2023-02-02","cited_by_count":25,"area":null},{"id":"2310.17087","title":"Good regularity creates large learning rate implicit biases: edge of stability, balancing, and catapult","created":"2023-10-26","cited_by_count":25,"area":null},{"id":"2405.21064","title":"Recurrent neural networks: vanishing and exploding gradients are not the end of the story","created":"2024-05-31","cited_by_count":25,"area":null},{"id":"2505.17863","title":"The emergence of sparse attention: impact of data distribution and benefits of repetition","created":"2025-05-23","cited_by_count":25,"area":null},{"id":"2302.08692","title":"SAM operates far from home: eigenvalue regularization as a dynamical phenomenon","created":"2023-02-17","cited_by_count":24,"area":null},{"id":"2410.10373","title":"Sharpness-Aware Minimization Efficiently Selects Flatter Minima Late in Training","created":"2024-10-14","cited_by_count":24,"area":null},{"id":"2402.17457","title":"Super Consistency of Neural Network Landscapes and Learning Rate Transfer","created":"2024-02-27","cited_by_count":23,"area":null},{"id":"2306.02601","title":"Aiming towards the minimizers: fast convergence of SGD for overparametrized problems","created":"2023-06-05","cited_by_count":22,"area":null},{"id":"2311.02076","title":"Universal Sharpness Dynamics in Neural Network Training: Fixed Point Analysis, Edge of Stability, and Route to Chaos","created":"2023-11-03","cited_by_count":22,"area":null},{"id":"2405.14578","title":"Surge Phenomenon in Optimal Learning Rate and Batch Size Scaling","created":"2024-05-23","cited_by_count":22,"area":null},{"id":"2406.06893","title":"Transformers Provably Learn Sparse Token Selection While Fully-Connected Nets Cannot","created":"2024-06-11","cited_by_count":22,"area":null},{"id":"2503.12811","title":"A Multi-Power Law for Loss Curve Prediction Across Learning Rate Schedules","created":"2025-03-17","cited_by_count":22,"area":null},{"id":"2507.02119","title":"Scaling Collapse Reveals Universal Dynamics in Compute-Optimally Trained Neural Networks","created":"2025-07-02","cited_by_count":22,"area":null},{"id":"2202.08788","title":"Global Convergence of Sub-gradient Method for Robust Matrix Recovery: Small Initialization, Noisy Measurements, and Over-parameterization","created":"2022-02-17","cited_by_count":21,"area":null},{"id":"2301.02312","title":"Training trajectories, mini-batch losses and the curious role of the learning rate","created":"2023-01-05","cited_by_count":21,"area":null},{"id":"2305.13064","title":"Gradient Descent Monotonically Decreases the Sharpness of Gradient Flow Solutions in Scalar Networks and Beyond","created":"2023-05-22","cited_by_count":21,"area":null},{"id":"2405.03676","title":"Why is SAM Robust to Label Noise?","created":"2024-05-06","cited_by_count":21,"area":null},{"id":"2106.15739","title":"On the Periodic Behavior of Neural Network Training with Batch Normalization and Weight Decay","created":"2021-06-29","cited_by_count":20,"area":null},{"id":"2306.09850","title":"Practical Sharpness-Aware Minimization Cannot Converge All the Way to Optima","created":"2023-06-16","cited_by_count":20,"area":null},{"id":"2107.09133","title":"The Limiting Dynamics of SGD: Modified Loss, Phase Space Oscillations, and Anomalous Diffusion","created":"2021-07-19","cited_by_count":19,"area":null},{"id":"2209.03695","title":"Training Scale-Invariant Neural Networks on the Sphere Can Happen in Three Regimes","created":"2022-09-08","cited_by_count":19,"area":null},{"id":"2302.12250","title":"Phase diagram of early training dynamics in deep neural networks: effect of the learning rate, depth, and width","created":"2023-02-23","cited_by_count":19,"area":null},{"id":"2306.09000","title":"When and Why Momentum Accelerates SGD:An Empirical Study","created":"2023-06-15","cited_by_count":19,"area":null},{"id":"2402.07193","title":"Parameter Symmetry and Noise Equilibrium of Stochastic Gradient Descent","created":"2024-02-11","cited_by_count":19,"area":null},{"id":"2502.02531","title":"Deep Linear Network Training Dynamics from Random Initialization: Data, Width, Depth, and Hyperparameter Transfer","created":"2025-02-04","cited_by_count":19,"area":null},{"id":"2404.19261","title":"High dimensional analysis reveals conservative sharpening and a stochastic edge of stability","created":"2024-04-30","cited_by_count":18,"area":null},{"id":"2205.15809","title":"Feature Learning in $L_{2}$-regularized DNNs: Attraction/Repulsion and Sparsity","created":"2022-05-31","cited_by_count":17,"area":null},{"id":"2305.19008","title":"Bottleneck Structure in Learned Features: Low-Dimension vs Regularity Tradeoff","created":"2023-05-30","cited_by_count":17,"area":null},{"id":"2309.12488","title":"Sharpness-Aware Minimization and the Edge of Stability","created":"2023-09-21","cited_by_count":17,"area":null},{"id":"2310.01687","title":"From Stability to Chaos: Analyzing Gradient Descent Dynamics in Quadratic Regression","created":"2023-10-02","cited_by_count":17,"area":null},{"id":"2403.04081","title":"Directional Smoothness and Gradient Methods: Convergence and Adaptivity","created":"2024-03-06","cited_by_count":17,"area":null},{"id":"2405.13456","title":"Deep linear networks for regression are implicitly regularized towards flat minima","created":"2024-05-22","cited_by_count":17,"area":null},{"id":"2405.16642","title":"Fast TRAC: A Parameter-Free Optimizer for Lifelong Reinforcement Learning","created":"2024-05-26","cited_by_count":17,"area":"deeprl"},{"id":"2406.08654","title":"Large Stepsize Gradient Descent for Non-Homogeneous Two-Layer Networks: Margin Improvement and Fast Optimization","created":"2024-06-12","cited_by_count":17,"area":null}],"next_cursor":"100"}