Sanjeev Arora

papers

1,827

total citations

papers (22)

Understanding the Generalization Benefit of Normalization Layers: Sharpness Reduction

NEURIPS 2022arXiv

citations

On the SDEs and Scaling Rules for Adaptive Gradient Algorithms

NEURIPS 2022arXiv

citations

Gradient Descent on Two-layer Nets: Margin Maximization and Simplicity Bias

NEURIPS 2021arXiv

citations

Goedel-Prover: A Frontier Model for Open-Source Automated Theorem Proving

COLM 2025arXiv

citations

Reconciling Modern Deep Learning with Traditional Optimization Analyses: The Intrinsic Learning Rate

NEURIPS 2020arXiv

citations

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

NEURIPS 2020arXiv

citations

Unintentional Unalignment: Likelihood Displacement in Direct Preference Optimization

ICLR 2025arXiv

citations

Implicit Bias of Gradient Descent on Reparametrized Models: On Equivalence to Mirror Descent

NEURIPS 2022arXiv

citations

Instruct-SkillMix: A Powerful Pipeline for LLM Instruction Tuning

ICLR 2025arXiv

citations

Language Models as Science Tutors

ICML 2024arXiv

citations

Trainable Transformer in Transformer

ICML 2024arXiv

citations

Generalizing from SIMPLE to HARD Visual Reasoning: Can We Mitigate Modality Imbalance in VLMs?

ICML 2025arXiv

citations

New Definitions and Evaluations for Saliency Methods: Staying Intrinsic, Complete and Sound

NEURIPS 2022arXiv

citations

On the Power of Context-Enhanced Learning in LLMs

ICML 2025arXiv

citations

Ineq-Comp: Benchmarking Human-Intuitive Compositional Reasoning in Automated Theorem Proving of Inequalities

NEURIPS 2025arXiv

citations

A Quadratic Synchronization Rule for Distributed Deep Learning

ICLR 2024arXiv

citations

AdaptMI: Adaptive Skill-based In-context Math Instructions for Small Language Models

COLM 2025arXiv

citations

Provable unlearning in topic modeling and downstream tasks

ICLR 2025arXiv

citations

Sanjeev Arora

papers (22)

LESS: Selecting Influential Data for Targeted Instruction Tuning

Evaluating Gradient Inversion Attacks and Defenses in Federated Learning

Fine-Tuning Language Models with Just Forward Passes

On the Validity of Modeling SGD with Stochastic Differential Equations (SDEs)

Understanding the Generalization Benefit of Normalization Layers: Sharpness Reduction

On the SDEs and Scaling Rules for Adaptive Gradient Algorithms

Gradient Descent on Two-layer Nets: Margin Maximization and Simplicity Bias

Goedel-Prover: A Frontier Model for Open-Source Automated Theorem Proving

Reconciling Modern Deep Learning with Traditional Optimization Analyses: The Intrinsic Learning Rate

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

Unintentional Unalignment: Likelihood Displacement in Direct Preference Optimization

Implicit Bias of Gradient Descent on Reparametrized Models: On Equivalence to Mirror Descent

Instruct-SkillMix: A Powerful Pipeline for LLM Instruction Tuning

Language Models as Science Tutors

Trainable Transformer in Transformer

Generalizing from SIMPLE to HARD Visual Reasoning: Can We Mitigate Modality Imbalance in VLMs?

New Definitions and Evaluations for Saliency Methods: Staying Intrinsic, Complete and Sound

On the Power of Context-Enhanced Learning in LLMs

Ineq-Comp: Benchmarking Human-Intuitive Compositional Reasoning in Automated Theorem Proving of Inequalities

A Quadratic Synchronization Rule for Distributed Deep Learning

AdaptMI: Adaptive Skill-based In-context Math Instructions for Small Language Models

Provable unlearning in topic modeling and downstream tasks

papers (22)

LESS: Selecting Influential Data for Targeted Instruction Tuning

Evaluating Gradient Inversion Attacks and Defenses in Federated Learning

Fine-Tuning Language Models with Just Forward Passes

On the Validity of Modeling SGD with Stochastic Differential Equations (SDEs)

Understanding the Generalization Benefit of Normalization Layers: Sharpness Reduction

On the SDEs and Scaling Rules for Adaptive Gradient Algorithms

Gradient Descent on Two-layer Nets: Margin Maximization and Simplicity Bias

Goedel-Prover: A Frontier Model for Open-Source Automated Theorem Proving

Reconciling Modern Deep Learning with Traditional Optimization Analyses: The Intrinsic Learning Rate

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

Unintentional Unalignment: Likelihood Displacement in Direct Preference Optimization

Implicit Bias of Gradient Descent on Reparametrized Models: On Equivalence to Mirror Descent

Instruct-SkillMix: A Powerful Pipeline for LLM Instruction Tuning

Language Models as Science Tutors

Trainable Transformer in Transformer

Generalizing from SIMPLE to HARD Visual Reasoning: Can We Mitigate Modality Imbalance in VLMs?

New Definitions and Evaluations for Saliency Methods: Staying Intrinsic, Complete and Sound

On the Power of Context-Enhanced Learning in LLMs

Ineq-Comp: Benchmarking Human-Intuitive Compositional Reasoning in Automated Theorem Proving of Inequalities

A Quadratic Synchronization Rule for Distributed Deep Learning

AdaptMI: Adaptive Skill-based In-context Math Instructions for Small Language Models

Provable unlearning in topic modeling and downstream tasks