Anvith Thudi

Published

• "Efficient Public Verification of Private ML via Regularization" Zoë Ruha Bell, Anvith Thudi, Olive Franzese-McLaughlin, Nicolas Papernot, Shafi Goldwasser. Proceedings of the 43rd International Conference on Machine Learning
• "Gauss-Newton Unlearning for the LLM Era" Lev McKinney, Anvith Thudi , Juhan Bae, Tara Rezaei Kheirkhah, Nicolas Papernot, Sheila A. McIlraith, Roger Baker Grosse. Proceedings of the 4th IEEE Conference on Secure and Trustworthy Machine Learning
• " k-Nearest Neighbour Adaptive Sampling (kNN-AS), a Simple Tool to Efficiently Explore Conformational Space" Evianne M. Rovers, Anvith Thudi, Jérôme Hénin, Chris Maddison, Matthieu Schapira. Journal of Chemical Theory and Computation
• "MixMin: Finding Data Mixtures via Convex Minimization"(Code) Anvith Thudi , Evianne Rovers, Yangjun Ruan, Tristan Thrush, Chris J. Maddison. Proceedings of the 42nd International Conference on Machine Learning
• "Leveraging Per-Instance Privacy for Machine Unlearning" Nazanin Mohammadi Sepahvand, Anvith Thudi , Berivan Isik, Ashmita Bhattacharyya, Nicolas Papernot, Eleni Triantafillou, Daniel M. Roy, Gintare Karolina Dziugaite. Proceedings of the 42nd International Conference on Machine Learning. Oral at TPDP workshop 2025
• "Fast Exact Unlearning for In-context Learning Data for LLMs" Andrei Muresanu, Anvith Thudi , Michael R. Zhang, Nicolas Papernot. Proceedings of the 42nd International Conference on Machine Learning
• "MixMax: Distributional Robustness in Function Space via Optimal Data Mixtures"(Code) Anvith Thudi , Chris J. Maddison. Proceedings of the 13th International Conference on Learning Representations
• "Selective Prediction via Neural Training Dynamics" Stephan Rabanser, Anvith Thudi , Kimia Hamidieh, Adam Dziedzic, Nicolas Papernot. Transactions on Machine Learning Research
• "Gradients Look Alike: Sensitivity is Often Overestimated in DP-SGD" (slides) (talk) Anvith Thudi , Hengrui Jia, Casey Meehan, Ilia Shumailov, Nicolas Papernot. Proceedings of the 33rd USENIX Security Symposium
• "Better Sparsifiers for Directed Eulerian Graphs" Sushant Sachdeva, Anvith Thudi , Yibin Zhao. Proceedings of the 51st EATCS International Colloquium on Automata, Languages and Programming
• "From Differential Privacy to Bounds on Membership Inference: Less can be More"Anvith Thudi, Ilia Shumailov, Franziska Boenisch, Nicolas Papernot. Transactions on Machine Learning Research
• "Training Private Models That Know What They Don't Know" Stephan Rabanser, Anvith Thudi , Abhradeep Thakurta, Krishnamurthy Dvijotham, Nicolas Papernot. Proceedings of the 37th Conference on Neural Information Processing Systems
• "Proof-of-Learning is Currently More Broken Than You Think" Congyu Fang, Hengrui Jia, Anvith Thudi, Mohammad Yaghini, Christopher A. Choquette-Choo, Natalie Dullerud, Varun Chandrasekaran, Nicolas Papernot. Proceedings of the 8th IEEE European Symposium on Security and Privacy
• "On the Necessity of Auditable Algorithmic Definitions for Machine Unlearning" (slides) (talk) Anvith Thudi, Hengrui Jia, Ilia Shumailov, Nicolas Papernot. Proceedings of the 31st USENIX Security Symposium
• "Unrolling SGD: Understanding Factors Influencing Machine Unlearning" (slides) Anvith Thudi, Gabriel Deza, Varun Chandrasekaran, Nicolas Papernot. Proceedings of the 7th IEEE European Symposium on Security and Privacy
• "Proof of Learning: Definitions and Practice"Hengrui Jia, Mohammad Yaghini, Christopher A. Choquette-Choo, Natalie Dullerud, Anvith Thudi, Varun Chandrasekaran, Nicolas Papernot. Proceedings of the 42nd IEEE Symposium on Security and Privacy

Preprints

• "Differentially Private Mixing of Public Datasets Improves Private Learning" Yufei Chen, Tejumade Afonja, Anvith Thudi, Nicolas Papernot
• "SoK: Machine Learning Governance" Varun Chandrasekaran, Hengrui Jia, Anvith Thudi, Adelin Travers, Mohammad Yaghini, Nicolas Papernot

Scaling data mixing for language and DNA pretraining

Scaling my work on data mixing to larger language models and DNA pretraining, with thousands of data sources, and mapping out the Pareto front of data mixing performance across multiple benchmarks. A collaboration with a wonderful team at Radical Numerics.

Arbitrating the integrity of stochastic gradient descent with proof-of-learning

An early approach to auditing the ownership of models by extending ideas from proof-of-work. The ideas from this paper eventually led to results on data forging, and subsequent unlearning work. A collaboration among students in the Cleverhans Lab.

Essays on C* Algebras

A collection of expository essays written while studying C* algebras and their classification under Prof. George Elliott during my undergrad. The goal of these essays was to present the main prerequisite ideas for a particular problem in one relatively short document (as a sort of reference guide with a common theme).

1. Category Theory and Classification
2. The Classification of "Nice" C* Algebras
3. The Atiyah-Singer Index Theorem (Note: indirectly motivated by a graduate course in Riemannian Geometry and the notion of index in terms of variations)
4. Graphs and C* Algebras (Note: was taking a graduate course in combinatorics and hence had graphs on my mind)
5. Irrational Rotational Algebras (Note: this is often the first concrete example on the classification of C* algebras one sees)