Chiyuan Zhang

Research Scientist@Google·OpenReview

研究方向

deep learning · generalization · memorization · overfitting · structured output learning · representation learning · speech recognition · manifold learning

Chiyuan Zhang

Private Hyperparameter Tuning with Ex-Post Guarantee

Scaling Embedding Layers in Language Models

Quantifying Cross-Modality Memorization in Vision-Language Models

Crosslingual Capabilities and Knowledge Barriers in Multilingual Large Language Models

URANIA: Differentially Private Insights into AI Use

Unlearn and Burn: Adversarial Machine Unlearning Requests Destroy Model Accuracy

Exploring and Mitigating Adversarial Manipulation of Voting-Based Leaderboards

MUSE: Machine Unlearning Six-Way Evaluation for Language Models

MATH-Perturb: Benchmarking LLMs' Math Reasoning Abilities against Hard Perturbations

Crosslingual Capabilities and Knowledge Barriers in Multilingual Large Language Models

Empirical Privacy Variance

Scaling Embedding Layers in Language Models

Scaling Laws for Differentially Private Language Models

On Memorization of Large Language Models in Logical Reasoning

Fantastic Copyrighted Beasts and How (Not) to Generate Them

Mind the Privacy Unit! User-Level Differential Privacy for Language Model Fine-Tuning

LabelDP-Pro: Learning with Label Differential Privacy via Projections

Scalable DP-SGD: Shuffling vs. Poisson Subsampling