Zhao Song

papers

1,169

total citations

papers (25)

Over-parameterized Adversarial Training: An Analysis Overcoming the Curse of Dimensionality

NEURIPS 2020arXiv

citations

Does Preprocessing Help Training Over-parameterized Neural Networks?

NEURIPS 2021arXiv

citations

On Computational Limits of Modern Hopfield Models: A Fine-Grained Complexity Analysis

ICML 2024arXiv

citations

Generalized Leverage Score Sampling for Neural Networks

NEURIPS 2020arXiv

citations

How to Protect Copyright Data in Optimization of Large Language Models?

AAAI 2024arXiv

citations

LazyDiT: Lazy Learning for the Acceleration of Diffusion Transformers

AAAI 2025arXiv

citations

InfoPrompt: Information-Theoretic Soft Prompt Tuning for Natural Language Understanding

NEURIPS 2023arXiv

citations

Low Rank Matrix Completion via Robust Alternating Minimization in Nearly Linear Time

ICLR 2024arXiv

citations

Algorithm and Hardness for Dynamic Attention Maintenance in Large Language Models

ICML 2024arXiv

citations

Breaking the Linear Iteration Cost Barrier for Some Well-known Conditional Gradient Methods Using MaxIP Data-structures

NEURIPS 2021arXiv

citations

Bypass Exponential Time Preprocessing: Fast Neural Network Training via Weight-Data Correlation Preprocessing

NEURIPS 2023arXiv

citations

Numerical Pruning for Efficient Autoregressive Models

AAAI 2025arXiv

citations

Dynamic Tensor Product Regression

NEURIPS 2022arXiv

citations

Unraveling the Smoothness Properties of Diffusion Models: A Gaussian Mixture Perspective

ICCV 2025arXiv

citations

Fundamental Limits of Prompt Tuning Transformers: Universality, Capacity and Efficiency

ICLR 2025arXiv

citations

Dissecting Submission Limit in Desk-Rejections: A Mathematical Analysis of Fairness in AI Conference Policies

ICML 2025arXiv

citations

Fast Distance Oracles for Any Symmetric Norm

NEURIPS 2022arXiv

citations

Fundamental Limits of Visual Autoregressive Transformers: Universal Approximation Abilities

ICML 2025

citations

Exact Representation of Sparse Networks with Symmetric Nonnegative Embeddings

NEURIPS 2023arXiv

citations

Binary Hypothesis Testing for Softmax Models and Leverage Score Models

ICML 2025arXiv

citations

H2O: Heavy-Hitter Oracle for Efficient Generative Inference of Large Language Models

NEURIPS 2023

citations

Differential Privacy for Euclidean Jordan Algebra with Applications to Private Symmetric Cone Programming

NEURIPS 2025arXiv

citations

Zhao Song

papers (25)

Evaluating Gradient Inversion Attacks and Defenses in Federated Learning

Scatterbrain: Unifying Sparse and Low-rank Attention

Fast Attention Requires Bounded Entries

Over-parameterized Adversarial Training: An Analysis Overcoming the Curse of Dimensionality

Does Preprocessing Help Training Over-parameterized Neural Networks?

On Computational Limits of Modern Hopfield Models: A Fine-Grained Complexity Analysis

Generalized Leverage Score Sampling for Neural Networks

How to Protect Copyright Data in Optimization of Large Language Models?

LazyDiT: Lazy Learning for the Acceleration of Diffusion Transformers

InfoPrompt: Information-Theoretic Soft Prompt Tuning for Natural Language Understanding

Low Rank Matrix Completion via Robust Alternating Minimization in Nearly Linear Time

Algorithm and Hardness for Dynamic Attention Maintenance in Large Language Models

Breaking the Linear Iteration Cost Barrier for Some Well-known Conditional Gradient Methods Using MaxIP Data-structures

Bypass Exponential Time Preprocessing: Fast Neural Network Training via Weight-Data Correlation Preprocessing

Numerical Pruning for Efficient Autoregressive Models

Dynamic Tensor Product Regression

Unraveling the Smoothness Properties of Diffusion Models: A Gaussian Mixture Perspective

Fundamental Limits of Prompt Tuning Transformers: Universality, Capacity and Efficiency

Dissecting Submission Limit in Desk-Rejections: A Mathematical Analysis of Fairness in AI Conference Policies

Fast Distance Oracles for Any Symmetric Norm

Fundamental Limits of Visual Autoregressive Transformers: Universal Approximation Abilities

Exact Representation of Sparse Networks with Symmetric Nonnegative Embeddings

Binary Hypothesis Testing for Softmax Models and Leverage Score Models

H2O: Heavy-Hitter Oracle for Efficient Generative Inference of Large Language Models

Differential Privacy for Euclidean Jordan Algebra with Applications to Private Symmetric Cone Programming

papers (25)

Evaluating Gradient Inversion Attacks and Defenses in Federated Learning

Scatterbrain: Unifying Sparse and Low-rank Attention

Fast Attention Requires Bounded Entries

Over-parameterized Adversarial Training: An Analysis Overcoming the Curse of Dimensionality

Does Preprocessing Help Training Over-parameterized Neural Networks?

On Computational Limits of Modern Hopfield Models: A Fine-Grained Complexity Analysis

Generalized Leverage Score Sampling for Neural Networks

How to Protect Copyright Data in Optimization of Large Language Models?

LazyDiT: Lazy Learning for the Acceleration of Diffusion Transformers

InfoPrompt: Information-Theoretic Soft Prompt Tuning for Natural Language Understanding

Low Rank Matrix Completion via Robust Alternating Minimization in Nearly Linear Time

Algorithm and Hardness for Dynamic Attention Maintenance in Large Language Models

Breaking the Linear Iteration Cost Barrier for Some Well-known Conditional Gradient Methods Using MaxIP Data-structures

Bypass Exponential Time Preprocessing: Fast Neural Network Training via Weight-Data Correlation Preprocessing

Numerical Pruning for Efficient Autoregressive Models

Dynamic Tensor Product Regression

Unraveling the Smoothness Properties of Diffusion Models: A Gaussian Mixture Perspective

Fundamental Limits of Prompt Tuning Transformers: Universality, Capacity and Efficiency

Dissecting Submission Limit in Desk-Rejections: A Mathematical Analysis of Fairness in AI Conference Policies

Fast Distance Oracles for Any Symmetric Norm

Fundamental Limits of Visual Autoregressive Transformers: Universal Approximation Abilities

Exact Representation of Sparse Networks with Symmetric Nonnegative Embeddings

Binary Hypothesis Testing for Softmax Models and Leverage Score Models

H2O: Heavy-Hitter Oracle for Efficient Generative Inference of Large Language Models

Differential Privacy for Euclidean Jordan Algebra with Applications to Private Symmetric Cone Programming