Mahdi Haghifam

Email(preferred): haghifam.mahdi@gmail.com Email: mhaghifam@ttic.edu

About Me

I am a machine learning researcher working on the foundations of machine learning. I like problems where theoretical understanding can shape how an algorithm is designed. My work spans privacy-preserving ML, memorization in learning systems, optimization, and generalization. More recently, I have been excited about the challenges of generative AI, and specifically principled approaches to inference-time algorithms for safety and efficiency. Recognition of my work includes a Research Fellowship at the Simons Institute (UC Berkeley), a Best Paper Award at ICML 2024, and multiple awards and scholarships from the University of Toronto for research excellence during my PhD studies.

I'm currently a Research Assistant Professor at TTIC. Previously, I was a Distinguished Postdoctoral Researcher at Northeastern University, working with Jonathan Ullman and Adam Smith. I completed my Ph.D. in Machine Learning (thesis) at the University of Toronto / Vector Institute, advised by Daniel M. Roy. I also hold B.Sc. and M.Sc. degrees in Electrical Engineering from Sharif University of Technology.

I've worked in industry at Google DeepMind (with Thomas Steinke) and at ServiceNow Research (with Gintare Karolina Dziugaite). See details here.

Selected Papers

• The Distillation Game in LLMs: Adaptive Attacks & Efficient Defenses
  Y. Allouah*, M. Haghifam*, R. Shokri, S. Koyejo
  Submitted, 2026
  Summary Slides Code

  Our main contribution is to cast distillation attack and defense as a minimax game: a model that must stay useful on one side, and on the other an attacker free to choose which of the released answers he trains on. Solving the attacker's side shows that antidistillation sampling (2025), the current state of the art, leaks about 50% more than its own evaluation reports. Solving the defender's side yields Product-of-Experts, a defense that holds up just as well at 1.6× generation time instead of 2.9×.

• Adaptive Generate-Rank-Verify: Inference-Time Search with Costly Verification
  S. Dughmi, M. Haghifam, Y.H. Kalayci (authors listed in alphabetic order)
  Submitted, 2026
  Summary Slides Code

  A common paradigm for inference-time compute is to generate many candidate responses, rank them with a cheap reward model, and pass the top-ranked ones to an expensive verifier. Since verification is often far more costly than generation and scoring, the policy faces a per-prompt compute allocation problem. We formalize this as active search: a sequential decision problem with a cheap reward-observation cost and an expensive verification cost. Our ADAP algorithm achieves a constant-factor competitive ratio against the distribution-aware optimum under a natural monotonicity assumption on the reward signal, and on HMMT math and LiveCodeBench coding it reaches 100% success rate at roughly 3x to 5x lower mean cost than the best fixed generate-and-verify policy. We also provide learning-theoretic view of this paradigm.

• The Sample Complexity of Membership Inference and Privacy Auditing
  M. Haghifam, A. Smith, J. Ullman
  FOCS 2026.
  Summary Slides

  We characterize the sample complexity of membership inference in the fundamental setting of Gaussian mean estimation. The main finding is that standard empirical attacks using O(n), where n is the number of training samples, reference samples can substantially underestimate the true privacy risk, since matching the fully-informed (Bayes-optimal) adversary requires far more reference data than the size of the training set itself, i.e., n^2.

• Information Complexity of Stochastic Convex Optimization: Applications to Generalization and Memorization
  I. Attias, G. K. Dziugaite, M. Haghifam, R. Livni, D. M. Roy (authors listed in alphabetic order)
  ICML 2024 (Oral, Best Paper Award: Top 10 of 10,000 submissions)
  Summary Talk

  We establish a strict tradeoff between excess loss and the conditional mutual information any algorithm extracts from its training data in stochastic convex optimization. Achieving small error necessarily requires revealing substantial information about individual samples, which yields lower bounds on memorization and a direct connection to vulnerability against membership inference attacks.

• Private Geometric Median
  M. Haghifam, T. Steinke, J. Ullman
  NeurIPS 2024
  Summary Talk Code

  We design a differentially private algorithm for the geometric median whose accuracy adapts to the effective diameter of the data, with no prior bound on scale or dispersion required. The utility guarantee improves on worst-case bounds whenever the data are concentrated, and the analysis avoids restrictive distributional assumptions.

• Faster Differentially Private Convex Optimization via Second-Order Methods
  A. Ganesh, M. Haghifam, T. Steinke, A. Thakurta (authors listed in alphabetic order)
  NeurIPS 2023
  Summary Code

  We show that second-order information can be used effectively under differential privacy, despite the added noise. Our algorithm, a private variant of the regularized cubic Newton method, adapts to the local shape of the loss and achieves the optimal excess loss for DP-ERM while running 10 to 40 times faster in wall-clock time than DP-SGD at comparable accuracy and privacy.

• Sharpened Generalization Bounds based on Conditional Mutual Information and an Application to Noisy, Iterative Algorithms
  M. Haghifam, J. Negrea, A. Khisti, D. M. Roy, G. K. Dziugaite
  NeurIPS 2020
  Summary Talk Code

  We give data-dependent generalization bounds for noisy iterative algorithms such as SGLD using the conditional mutual information framework. Unlike worst-case Lipschitz- or norm-based bounds, the resulting estimates depend on the algorithm's actual trajectory on the training data and remain non-vacuous in modern overparameterized settings.

Industry Internship Experience

Professional Service

Area Chair: ALT (2026), ICLR (2027), NeurIPS (2026)  ·  Senior Program Committee: ALT (2027)  ·  Program Committee: SaTML (2024, 2026), COLT (2026)

Contact Me!

Feel free to reach out if you'd like to discuss research ideas. Also, I'm happy to offer guidance and support to those applying to graduate programs, especially individuals who might not typically have access to such assistance