I am a Postdoc at Harvard University where my research is focused on trustworthy AI including topics such as unlearning, robustness, privacy and interpretability. There, I am working with Himabindu Lakkaraju and Seth Neel. Before joining Harvard University, I completed my PhD at the University of Tübingen with Gjergji Kasneci.
Recent News
I am on the 2024/2025 job market. Feel free to contact me if there is an opportunity that fits.
June'25: Spotlight Talk at ICML'24 workshop: Generalizing Trust: Weak-to-Strong Trustworthiness in Language Models
May '25: New preprint coming soon: Efficiently Verifiable Proofs of Data Attribution
Jan '25: Paper accepted at ICLR'25: Machine Unlearning Fails to Remove Data Poisoning Attacks
Jan '25: New paper on Arxiv: Generalizing Trust: Weak-to-Strong Trustworthiness in Language Models
Dec '24: ICLR'25 Workshop accepted @ICLR'25: Trust in LLMs & LLM Applications
July '24: Presenting In-Context Unlearning @ICML'24 in Vienna
June '24: Spotlight Talk at ICML'24 GenLaw workshop: Machine Unlearning Fails to Remove Data Poisoning Attacks
May '24: Paper accepted at ICML'24: In-Context Unlearning: Language Models as Few Shot Unlearners
Research Directions
Ensuring that ML models are aligned with human values and norms seems like one of the most critical problems we can be working on now. The specific directions I am interested in include:
Fundamental tradeoffs between trustworthiness properties: My work systematically reveals that counterfactual explanation methods, which are designed to enhance interpretability, can inadvertently compromise robustness and privacy:
Probabilistically Robust Recourse: Navigating the Trade-offs between Costs and Robustness in Algorithmic Recourse, ICLR 2023;
On the Trade-Off between Actionable Explanations and the Right to be Forgotten, ICLR 2023;
On the Privacy Risks of Algorithmic Recourse, AISTATS 2023;
Exploring Counterfactual Explanations Through the Lens of Adversarial Attacks, AISTATS 2022;
On Counterfactual Explanations under Predictive Multiplicity, UAI 2020;
Learning Model-Agnostic Counterfactual Explanations, WWW 2020.
Privacy and unlearning: My work has developed novel privacy frameworks and unlearning methods that demonstrate that we may achieve high utility while simultaneously ensuring data confidentiality:
Gaussian Membership Inference Privacy, Neurips 2023;
Language Models are Realistic Tabular Data Generators, ICLR 2023;
I Prefer not to Say - Protecting User Consent in Models with Optional Personal Data, AAAI 2024;
In-Context Unlearning: Language Models as Few Shot Unlearners, ICML 2024;
Machine Unlearning Fails to Remove Data Poisoning Attacks, ICLR 2025.
Reliable ML evaluations: I have devised large scale evaluations that test state-of-the-art ML methods:
OpenXAI: Towards Transparent Evaluations of Model Explanations, NeurIPS 2022;
CARLA: A Counterfactual Explanation Benchmark, NeurIPS 2021.
Generalizing Trust: Weak-to-Strong Trustworthiness in Language Models, Preprint 2025.
Efficiently Verifiable Proofs of Data Attribution, Preprint 2025.
Selected Publications
Preprint
Efficiently Verifiable Proofs of Data Attribution, coming soon
Ari Karchmer, Seth Neel, Martin Pawelczyk
Tags: Data Attributions, Auditing
Preprint
Generalizing Trust: Weak-to-Strong Trustworthiness in Language Models , [arxiv]
Martin Pawelczyk, Aounon Kumar, Lillian Sun, Zhenting Qi, Himabindu Lakkaraju
Tags: LLMs, OOD & Adv Robustness, Fairness, Privacy
ICLR 2023
ICLR 2023
On the Trade-Off between Actionable Explanations and the Right to be Forgotten, [arxiv]
Martin Pawelczyk, Tobias Leemann, Asia Biega*, and Gjergji Kasneci* (*equal contribution)
Tags: Counterfactual Explanations, Unlearning, Adversarial Attacks, Robustness, Theory
NeurIPS 2022
OpenXAI: Towards a Transparent Evaluation of Model Explanations, [arxiv] [github] [website], Spotlight @ ICLR 2022 PAIR^2Struc Workshop
Chirag Agarwal, Eshika Saxena, Satyapriya Krishna, Martin Pawelczyk, Nari Johnson, Isha Puri, Marinka Zitnik, Himabindu Lakkaraju
Tags: Feature Attributions, Evaluation, Software
AISTATS 2022
NeurIPS 2021
CARLA - A Python Library to Benchmark Counterfactual Explanations and Algorithmic Recourse Methods, [arXiv] [documentation] [github] [slides] [poster], Spotlight @ ICML 2022 Algorithmic Recourse Workshop
Martin Pawelczyk, Sascha Bielawski, Johan van den Heuvel, Tobias Richter* and Gjergji Kasneci* (*equal contribution)
Tags: Counterfactual Explanations, Evaluation, Software
UAI 2020
On Learning Counterfactual Explanations under Predictive Multiplicity, [arXiv]
Martin Pawelczyk, Klaus Broelemann and Gjergji Kasneci
Tags: Counterfactual Explanations, Robustness