Publications

Memorization Sinks: Isolating Memorization during LLM Training

Gaurav R. Ghosal, Pratyush Maini, Aditi Raghunathan

Unlocking Post-hoc Dataset Inference with Synthetic Data

Bihe Zhao, Pratyush Maini, Franziska Boenisch, Adam Dziedzic

STAMP Your Content: Proving Dataset Membership via Watermarked Rephrasings

Saksham Rastogi, Pratyush Maini, Danish Pruthi

OpenUnlearning: Accelerating LLM Unlearning via Unified Benchmarking of Methods and Metrics

Vineeth Dorna, Anmol Mekala, Wenlong Zhao, Andrew McCallum, Zachary C. Lipton, J. Zico Kolter, Pratyush Maini

MAGIC: Diffusion Model Memorization Auditing via Generative Image Compression

Gunjan Dhanuka, Sumukh K. Aithal, Avi Schwarzschild, Zhili Feng, J. Zico Kolter, Zachary C. Lipton, Pratyush Maini

Reassessing EMNLP 2024's Best Paper: Does Divergence-Based Calibration for MIAs Hold Up?

Pratyush Maini, Anshuman Suri

LLM Dataset Inference: Did you train on my dataset?

Pratyush Maini*, Hengrui Jia*, Nicolas Papernot, Adam Dziedzic

Black-box detection of whether a dataset was used to train an LLM.

Rethinking LLM Memorization through the Lens of Adversarial Compression

Avi Schwarzschild, Zhili Feng, Pratyush Maini, Zachary C. Lipton, J. Zico Kolter

Understanding Hallucinations in Diffusion Models through Mode Interpolation

Sumukh K. Aithal, Pratyush Maini, Zachary C. Lipton, J. Zico Kolter

TOFU: A Task of Fictitious Unlearning for LLMs

Pratyush Maini*, Zhili Feng*, Avi Schwarzschild*, Zachary C. Lipton, J. Zico Kolter

Benchmarking machine unlearning methods for large language models.

Can Neural Network Memorization Be Localized?

Pratyush Maini, Michael C. Mozer, Hanie Sedghi, Zachary C. Lipton, J. Zico Kolter, Chiyuan Zhang

Individual neurons and layers do not solely determine what a neural network memorizes.

Dataset Inference: Ownership Resolution in Machine Learning

Pratyush Maini, Mohammad Yaghini, Nicolas Papernot

First work on dataset inference: determining if a dataset was used for training.

BeyondWeb: Lessons from Scaling Synthetic Data for Trillion-scale Pretraining

Pratyush Maini, Vineeth Dorna, Parth Doshi, Aldo Carranza, Fan Pan, Jack Urbanek, Paul Burstein, Alex Fang, Alvin Deng, Amro Abbas, Brett Larsen, Cody Blakeney, Charvi Bannur, Christina Baek, Darren Teh, David Schwab, Haakon Mongstad, Haoli Yin, Josh Wills, Kaleigh Mentzer, Luke Merrick, Ricardo Monti, Rishabh Adiga, Siddharth Joshi, Spandan Das, Zhengping Wang, Bogdan Gaza, Ari Morcos, Matthew Leavitt

Safety Pretraining: Toward the Next Generation of Safe AI

Pratyush Maini*, Sachin Goyal*, Dylan Sam*, Alex Robey, Yash Savani, Yiding Jiang, Andy Zou, Matt Fredrikson, Zachary C. Lipton, J. Zico Kolter

A framework for training AI systems to be inherently safer through specialized pretraining.

Peeking Behind Closed Doors: Risks of LLM Evaluation by Private Data Curators

Harshay Bansal, Pratyush Maini

Rephrasing the Web: A Recipe for Compute and Data-Efficient Language Modeling

Pratyush Maini, Skyler Seto, He Bai, David Grangier, Yizhe Zhang, Navdeep Jaitly

Paraphrasing web data into Q/A pairs significantly improves language model training efficiency.

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

Sachin Goyal, Pratyush Maini, Zachary C. Lipton, Aditi Raghunathan, J. Zico Kolter

Training compute-optimal models requires data filtering that scales with available compute.

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

Pratyush Maini, Sachin Goyal, Zachary C. Lipton, J. Zico Kolter, Aditi Raghunathan

Characterizing Datapoints via Second-Split Forgetting

Pratyush Maini, Saurabh Garg, Zachary C. Lipton, J. Zico Kolter

Model-tuning via Prompts Makes NLP Models Adversarially Robust

Mrigank Raman, Pratyush Maini, J. Zico Kolter, Zachary C. Lipton, Danish Pruthi

Perturbation Type Categorization for Multiple l_p Bounded Adversarial Robustness

Pratyush Maini, Xinyun Chen, Bo Li, Dawn Song

Adversarial Robustness Against the Union of Multiple Perturbation Models

Pratyush Maini, Eric Wong, J. Zico Kolter

Data-free Model Extraction

Jean-Baptiste Truong*, Pratyush Maini*, Robert J. Walls, Nicolas Papernot

Why and When Should You Pool? Analyzing Pooling in Recurrent Architectures

Pratyush Maini, Keshav Kolluru, Danish Pruthi, Mausam

BeyondWeb: Lessons from Scaling Synthetic Data for Trillion-scale Pretraining

Pratyush Maini, Vineeth Dorna, Parth Doshi, et al.

Safety Pretraining: Toward the Next Generation of Safe AI

Pratyush Maini*, Sachin Goyal*, Dylan Sam*, et al.

Memorization Sinks: Isolating Memorization during LLM Training

Gaurav R. Ghosal, Pratyush Maini, Aditi Raghunathan

Unlocking Post-hoc Dataset Inference with Synthetic Data

Bihe Zhao, Pratyush Maini, Franziska Boenisch, Adam Dziedzic

STAMP Your Content: Proving Dataset Membership via Watermarked Rephrasings

Saksham Rastogi, Pratyush Maini, Danish Pruthi

OpenUnlearning: Accelerating LLM Unlearning via Unified Benchmarking

Vineeth Dorna, Anmol Mekala, et al., Pratyush Maini

MAGIC: Diffusion Model Memorization Auditing via Generative Image Compression

Gunjan Dhanuka, Sumukh K. Aithal, et al., Pratyush Maini

Reassessing EMNLP 2024's Best Paper

Pratyush Maini, Anshuman Suri

Peeking Behind Closed Doors: Risks of LLM Evaluation by Private Data Curators

Harshay Bansal, Pratyush Maini

LLM Dataset Inference: Did you train on my dataset?

Pratyush Maini*, Hengrui Jia*, Nicolas Papernot, Adam Dziedzic

Rethinking LLM Memorization through the Lens of Adversarial Compression

Avi Schwarzschild, Zhili Feng, Pratyush Maini, Zachary C. Lipton, J. Zico Kolter

Understanding Hallucinations in Diffusion Models through Mode Interpolation

Sumukh K. Aithal, Pratyush Maini, Zachary C. Lipton, J. Zico Kolter

TOFU: A Task of Fictitious Unlearning for LLMs

Pratyush Maini*, Zhili Feng*, Avi Schwarzschild*, Zachary C. Lipton, J. Zico Kolter

Rephrasing the Web: A Recipe for Compute and Data-Efficient Language Modeling

Pratyush Maini, Skyler Seto, He Bai, David Grangier, Yizhe Zhang, Navdeep Jaitly

Scaling Laws for Data Filtering

Sachin Goyal, Pratyush Maini, Zachary C. Lipton, Aditi Raghunathan, J. Zico Kolter

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

Pratyush Maini, Sachin Goyal, Zachary C. Lipton, J. Zico Kolter, Aditi Raghunathan

Can Neural Network Memorization Be Localized?

Pratyush Maini, Michael C. Mozer, Hanie Sedghi, Zachary C. Lipton, J. Zico Kolter, Chiyuan Zhang

Model-tuning via Prompts Makes NLP Models Adversarially Robust

Mrigank Raman, Pratyush Maini, J. Zico Kolter, Zachary C. Lipton, Danish Pruthi

Characterizing Datapoints via Second-Split Forgetting

Pratyush Maini, Saurabh Garg, Zachary C. Lipton, J. Zico Kolter

Perturbation Type Categorization for Multiple l_p Bounded Adversarial Robustness

Pratyush Maini, Xinyun Chen, Bo Li, Dawn Song

Dataset Inference: Ownership Resolution in Machine Learning

Pratyush Maini, Mohammad Yaghini, Nicolas Papernot

Data-free Model Extraction

Jean-Baptiste Truong*, Pratyush Maini*, Robert J. Walls, Nicolas Papernot

Adversarial Robustness Against the Union of Multiple Perturbation Models

Pratyush Maini, Eric Wong, J. Zico Kolter

Why and When Should You Pool? Analyzing Pooling in Recurrent Architectures

Pratyush Maini, Keshav Kolluru, Danish Pruthi, Mausam