Chiyuan Zhang

papers

3,249

total citations

papers (20)

Do Vision Transformers See Like Convolutional Neural Networks?

NEURIPS 2021arXiv

1,247

citations

What is being transferred in transfer learning?

NEURIPS 2020arXiv

592

citations

What Neural Networks Memorize and Why: Discovering the Long Tail via Influence Estimation

NEURIPS 2020arXiv

575

citations

Understanding and Improving Robustness of Vision Transformers through Patch-based Negative Augmentation

NEURIPS 2022arXiv

citations

Fantastic Copyrighted Beasts and How (Not) to Generate Them

ICLR 2025arXiv

citations

How Private are DP-SGD Implementations?

ICML 2024arXiv

citations

Learning to Reason with Neural Networks: Generalization, Unseen Data and Boolean Measures

NEURIPS 2022arXiv

citations

User-Level Differential Privacy With Few Examples Per User

NEURIPS 2023arXiv

citations

Exploring and Mitigating Adversarial Manipulation of Voting-Based Leaderboards

ICML 2025arXiv

citations

Scaling Laws for Differentially Private Language Models

ICML 2025arXiv

citations

Scaling Embedding Layers in Language Models

NEURIPS 2025arXiv

citations

Crosslingual Capabilities and Knowledge Barriers in Multilingual Large Language Models

COLM 2025arXiv

citations

Sparsity-Preserving Differentially Private Training of Large Embedding Models

NEURIPS 2023arXiv

citations

Empirical Privacy Variance

ICML 2025arXiv

citations

URANIA: Differentially Private Insights into AI Use

COLM 2025arXiv

citations

Optimal Unbiased Randomizers for Regression with Label Differential Privacy

NEURIPS 2023

citations

Chiyuan Zhang

papers (20)

Do Vision Transformers See Like Convolutional Neural Networks?

What is being transferred in transfer learning?

What Neural Networks Memorize and Why: Discovering the Long Tail via Influence Estimation

Counterfactual Memorization in Neural Language Models

Deep Learning with Label Differential Privacy

MUSE: Machine Unlearning Six-Way Evaluation for Language Models

The Privacy Onion Effect: Memorization is Relative

Understanding and Improving Robustness of Vision Transformers through Patch-based Negative Augmentation

Fantastic Copyrighted Beasts and How (Not) to Generate Them

How Private are DP-SGD Implementations?

Learning to Reason with Neural Networks: Generalization, Unseen Data and Boolean Measures

User-Level Differential Privacy With Few Examples Per User

Exploring and Mitigating Adversarial Manipulation of Voting-Based Leaderboards

Scaling Laws for Differentially Private Language Models

Scaling Embedding Layers in Language Models

Crosslingual Capabilities and Knowledge Barriers in Multilingual Large Language Models

Sparsity-Preserving Differentially Private Training of Large Embedding Models

Empirical Privacy Variance

URANIA: Differentially Private Insights into AI Use

Optimal Unbiased Randomizers for Regression with Label Differential Privacy

papers (20)

Do Vision Transformers See Like Convolutional Neural Networks?

What is being transferred in transfer learning?

What Neural Networks Memorize and Why: Discovering the Long Tail via Influence Estimation

Counterfactual Memorization in Neural Language Models

Deep Learning with Label Differential Privacy

MUSE: Machine Unlearning Six-Way Evaluation for Language Models

The Privacy Onion Effect: Memorization is Relative

Understanding and Improving Robustness of Vision Transformers through Patch-based Negative Augmentation

Fantastic Copyrighted Beasts and How (Not) to Generate Them

How Private are DP-SGD Implementations?

Learning to Reason with Neural Networks: Generalization, Unseen Data and Boolean Measures

User-Level Differential Privacy With Few Examples Per User

Exploring and Mitigating Adversarial Manipulation of Voting-Based Leaderboards

Scaling Laws for Differentially Private Language Models

Scaling Embedding Layers in Language Models

Crosslingual Capabilities and Knowledge Barriers in Multilingual Large Language Models

Sparsity-Preserving Differentially Private Training of Large Embedding Models

Empirical Privacy Variance

URANIA: Differentially Private Insights into AI Use

Optimal Unbiased Randomizers for Regression with Label Differential Privacy