Aditi Raghunathan

papers

1,325

total citations

papers (19)

Agreement-on-the-line: Predicting the Performance of Neural Networks under Distribution Shift

NEURIPS 2022arXiv

108

citations

Enabling certification of verification-agnostic networks via memory-efficient semidefinite programming

NEURIPS 2020arXiv

101

citations

Dissecting Adversarial Robustness of Multimodal LM Agents

ICLR 2025arXiv

citations

Scaling Laws for Data Filtering— Data Curation cannot be Compute Agnostic

CVPR 2024arXiv

citations

Repetition Improves Language Model Embeddings

ICLR 2025arXiv

citations

T-MARS: Improving Visual Representations by Circumventing Text Feature Learning

ICLR 2024arXiv

citations

Overtrained Language Models Are Harder to Fine-Tune

ICML 2025arXiv

citations

Understanding Finetuning for Factual Knowledge Extraction

ICML 2024arXiv

citations

Roll the dice & look before you leap: Going beyond the creative limits of next-token prediction

ICML 2025arXiv

citations

Complementary Benefits of Contrastive Learning and Self-Training Under Distribution Shift

NEURIPS 2023arXiv

citations

Reasoning as an Adaptive Defense for Safety

NEURIPS 2025arXiv

citations

Sharpness-Aware Minimization Enhances Feature Quality via Balanced Learning

ICLR 2024arXiv

citations

Context-Parametric Inversion: Why Instruction Finetuning May Not Actually Improve Context Reliance

ICLR 2025

citations

Memorization Sinks: Isolating Memorization during LLM Training

ICML 2025arXiv

citations

Mitigating Modal Imbalance in Multimodal Reasoning

COLM 2025arXiv

citations

Test Time Adaptation via Conjugate Pseudo-labels

NEURIPS 2022

citations

Aditi Raghunathan

papers (19)

The Pitfalls of Simplicity Bias in Neural Networks

Finetune Like You Pretrain: Improved Finetuning of Zero-Shot Vision Models

Understanding Catastrophic Forgetting in Language Models via Implicit Inference

Agreement-on-the-line: Predicting the Performance of Neural Networks under Distribution Shift

Enabling certification of verification-agnostic networks via memory-efficient semidefinite programming

Dissecting Adversarial Robustness of Multimodal LM Agents

Scaling Laws for Data Filtering— Data Curation cannot be Compute Agnostic

Repetition Improves Language Model Embeddings

T-MARS: Improving Visual Representations by Circumventing Text Feature Learning

Overtrained Language Models Are Harder to Fine-Tune

Understanding Finetuning for Factual Knowledge Extraction

Roll the dice & look before you leap: Going beyond the creative limits of next-token prediction

Complementary Benefits of Contrastive Learning and Self-Training Under Distribution Shift

Reasoning as an Adaptive Defense for Safety

Sharpness-Aware Minimization Enhances Feature Quality via Balanced Learning

Context-Parametric Inversion: Why Instruction Finetuning May Not Actually Improve Context Reliance

Memorization Sinks: Isolating Memorization during LLM Training

Mitigating Modal Imbalance in Multimodal Reasoning

Test Time Adaptation via Conjugate Pseudo-labels

papers (19)

The Pitfalls of Simplicity Bias in Neural Networks

Finetune Like You Pretrain: Improved Finetuning of Zero-Shot Vision Models

Understanding Catastrophic Forgetting in Language Models via Implicit Inference

Agreement-on-the-line: Predicting the Performance of Neural Networks under Distribution Shift

Enabling certification of verification-agnostic networks via memory-efficient semidefinite programming

Dissecting Adversarial Robustness of Multimodal LM Agents

Scaling Laws for Data Filtering— Data Curation cannot be Compute Agnostic

Repetition Improves Language Model Embeddings

T-MARS: Improving Visual Representations by Circumventing Text Feature Learning

Overtrained Language Models Are Harder to Fine-Tune

Understanding Finetuning for Factual Knowledge Extraction

Roll the dice & look before you leap: Going beyond the creative limits of next-token prediction

Complementary Benefits of Contrastive Learning and Self-Training Under Distribution Shift

Reasoning as an Adaptive Defense for Safety

Sharpness-Aware Minimization Enhances Feature Quality via Balanced Learning

Context-Parametric Inversion: Why Instruction Finetuning May Not Actually Improve Context Reliance

Memorization Sinks: Isolating Memorization during LLM Training

Mitigating Modal Imbalance in Multimodal Reasoning

Test Time Adaptation via Conjugate Pseudo-labels