I am a Senior Researcher at Microsoft Research in New York City, working on AI as part of AI Frontiers. Previously, I was an IFML Postdoctoral Fellow at UT Austin, hosted by Adam Klivans and Raghu Meka. I received my PhD in Computer Science from the University of Wisconsin-Madison, where I was advised by Christos Tzamos. Prior to UW-Madison, I studied Electrical and Computer Engineering at the National Technical University of Athens, where I was advised by Dimitris Fotakis.

Here is my CV.

News

• Apr, 2026: Excited to release Memento — we extend the effective output length of LLMs by splitting chain-of-thought into blocks and summaries, then evicting old blocks from the KV cache so the model can reason longer within a fixed context window. Check out the blogpost, the code, the paper, and the OpenMementos dataset!

• Jul, 2025: I joined Microsoft Research in New York City as a Senior Researcher.

• Jul, 2025: Our work Train for the Worst, Plan for the Best: Understanding Token Ordering in Masked Diffusions received an Outstanding Paper Award at ICML 2025.

• May, 2025: New papers accepted to ICML 2025, COLT 2025, and ICLR 2025.

• Jul, 2024: Our work Smoothed Analysis for Learning Concepts with Low Intrinsic Dimension received the Best Paper Award at COLT 2024.

Publications

1. Train for the Worst, Plan for the Best: Understanding Token Ordering in Masked Diffusions
   w/ J. Kim, K. Shah, S. M. Kakade, S. Chen
   Outstanding Paper Award
   ICML 2025

2. Learning General Gaussian Mixtures with Efficient Score Matching
   w/ S. Chen, K. Shah
   COLT 2025

3. Online Linear Classification with Massart Noise
   w/ I. Diakonikolas, C. Tzamos, N. Zarifis
   ICML 2025

4. Oracle-Efficient Truncated Statistics
   w/ K. Karatapanis, C. Tzamos
   ICLR 2025

5. Smoothed Analysis for Learning Concepts with Low Intrinsic Dimension
   w/ G. Chandrasekaran, A. Klivans, R. Meka, K. Stavropoulos
   Best Paper Award COLT 2024
   IPAM 2024 Long Talk Video

6. Active Learning with Simple Questions
   w/ M. Ma, C. Tzamos
   COLT 2024

7. Agnostically Learning Multi-index Models with Queries
   w/ I. Diakonikolas, D. Kane, C. Tzamos, N. Zarifis
   FOCS 2024

8. Super Non-singular Decompositions of Polynomials and their
   Application to Robustly Learning Low-degree PTFs
   w/ I. Diakonikolas, D M. Kane, S. Liu, N. Zarifis
   STOC 2024

9. Efficient Discrepancy Testing for Learning with Distribution Shift
   w/ G. Chandrasekaran, A. Klivans, K. Stavropoulos, A. Vasilyan
   NeurIPS 2024

10. Active Classification with Few Queries under Misspecification
    w/ C. Tzamos, M. Ma
    Spotlight Presentation
    NeurIPS 2024

11. Learning Noisy Halfspaces with a Margin: Massart is no Harder than Random
    w/ G. Chandrasekaran, K. Stavropoulos, K. Tian
    Spotlight Presentation
    NeurIPS 2024

12. Optimizing Solution-Samplers for Combinatorial Problems:
    The Landscape of Policy Gradient Methods
    w/ C. Caramanis, D. Fotakis, A. Kalavasis, C. Tzamos
    Selected for Oral Presentation
    NeurIPS 2023

13. SLaM: Student-Label Mixing for Distillation with Unlabeled Examples
    w/ F. Iliopoulos, K. Trinh, C. Baykal, G. Menghani, V. Erik
    NeurIPS 2023

14. The Gain from Ordering in Online Learning
    w/ M. Ma, C. Tzamos
    NeurIPS 2023

15. Efficient Testable Learning of Halfspaces with Adversarial Label Noise
    w/ I. Diakonikolas, D M. Kane, S. Liu, N. Zarifis
    NeurIPS 2023

16. Self Directed Linear Classification
    w/ I. Diakonikolas, C. Tzamos, N. Zarifis
    COLT 2023

17. Weighted Distillation with Unlabeled Examples
    w/ F. Iliopoulos, C. Baykal, G. Menghani, K. Trinh, V. Erik
    NeurIPS 2022

18. Linear Label Ranking with Bounded Noise
    w/ D. Fotakis, A. Kalavasis, C. Tzamos
    Selected for Oral Presentation
    NeurIPS 2022

19. Learning General Halfspaces with Adversarial Label Noise via Online Gradient Descent
    w/ I. Diakonikolas, C. Tzamos, N. Zarifis
    ICML 2022

20. Learning a Single Neuron with Adversarial Label Noise via Gradient Descent
    w/ I. Diakonikolas, C. Tzamos, N. Zarifis
    COLT 2022

21. Learning General Halfspaces with General Massart Noise under the Gaussian Distribution
    w/ I. Diakonikolas, D. Kane, C. Tzamos, N. Zarifis
    STOC 2022

22. A Statistical Taylor Theorem and Extrapolation of Truncated Densities
    w/ C. Daskalakis, C. Tzamos, M. Zampetakis
    COLT 2021

23. Agnostic Proper Learning of Halfspaces under Gaussian Marginals
    w/ I. Diakonikolas, D. Kane, C. Tzamos, N. Zarifis
    COLT 2021

24. Efficient Algorithms for Learning from Coarse Labels
    w/ D. Fotakis, A. Kalavasis, C. Tzamos
    COLT 2021

25. Learning Online Algorithms with Distributional Advice
    w/ I. Diakonikolas, C. Tzamos, A. Vakilian, N. Zarifis
    ICML 2021

26. A Polynomial Time Algorithm For Learning Halfspaces with Tsybakov Noise
    w/ I. Diakonikolas, D. Kane, C. Tzamos, N. Zarifis
    STOC 2021

27. Learning Halfspaces with Tsybakov Noise
    w/ I. Diakonikolas, C. Tzamos, N. Zarifis
    STOC 2021
    Conference version merged with the above paper

28. Non-Convex SGD Learns Halfspaces with Adversarial Label Noise
    w/ I. Diakonikolas, C. Tzamos, N. Zarifis
    NeurIPS 2020

29. Learning Halfspaces with Massart Noise Under Structured Distributions
    w/ I. Diakonikolas, C. Tzamos, N. Zarifis
    COLT 2020

30. Algorithms and SQ Lower Bounds for PAC Learning One-Hidden-Layer ReLU Networks
    w/ I. Diakonikolas, D. Kane, N. Zarifis
    COLT 2020

31. Efficient Truncated Statistics with Unknown Truncation
    w/ C. Tzamos, M. Zampetakis
    FOCS 2019

32. Removing Bias in Maching Learning via Truncated Statistics
    w/ C. Daskalakis, C. Tzamos, M. Zampetakis
    Manuscript

33. Opinion Dynamics with Limited Information
    w/ D. Fotakis, V. Kandiros, S. Skoulakis
    WINE 2018

34. Learning Powers of Poisson Binomial Distributions
    w/ D. Fotakis, P. Krysta, P. Spirakis
    Manuscript

Service

• Program Committees: ITCS-2025

• Reviewer: FOCS (2020, 2021, 2023) , STOC (2020, 2024), COLT (2023), SODA (2019), NeurIPS (2021, 2023), WINE(2018), ICML(2023, 2021, 2020), EC (2022, 2020), MFCS (2018), TCS (2018, 2021), ALT (2021)

Talks

• Smoothed Analysis for Learning Concepts with Low Intrinsic Dimension
  1. Best Paper Award Talk, COLT 2024
  2. EnCORE Workshop on Computational vs Statistical Gaps in Learning and Optimization, 2024 UCLA IPAM 2024
  3. Theory Seminar, 2024, University of Southern California

• Optimizing Solution-Samplers for Combinatorial Problems, NeurIPS 2023 Oral

• SLaM: Student-Label Mixing for Distillation with Unlabeled Examples, NeurIPS 2023

• Learning General Halfspaces with General Massart Noise, STOC 2022

• A Statistical Taylor’s Theorem and Extrapolation of Truncated Densities, COLT 2021

• Agnostic Proper Learning of Halfspaces under Gaussian Marginals COLT 2021

• Efficient Algorithms for Learning Halfspaces with Tsybakov Noise, STOC 2021

• Non-Convex SGD Learns Halfspaces with Adversarial Label Noise, NeurIPS 2020

• Learning Halfspaces with Massart Noise Under Structured Distributions, COLT 2020

• Efficient Truncated Statistics with Unknown Truncation, FOCS 2019, Video

• Learning PBD Powers, ECCO Research Seminar 2017, University of Liverpool

• Learning Theory Study Group, Corelab NTUA, 2017
  1. Intoduction, No Free Lunch, Bias-Variance Tradeoff
  2. VC-Dimension
• Convex Optimization Minicourse, Corelab NTUA, 2017
  1. Convex Problems, LP
  2. QP, SOCP
  3. SDP, GW MaxCut
  4. Vector Optimization, Duality
• Programming with Dependent Types, NTUA, 2015