Prof. Martin Vechev

Research areas: Programming Languages and Artificial Intelligence
Google Scholar: link


Martin Vechev is originally from Sofia, Bulgaria where he was born and grew up. He is a Full Professor of Computer Science at ETH Zurich where he leads the Secure, Reliable, and Intelligent Systems Lab and a founder and architect of INSAIT.


Prior to ETH Zurich, Vechev was a Research Staff Member at the IBM T.J. Watson Research Center in New York, USA (2007-2011). He obtained his Ph.D. from the University of Cambridge, England (2003-2008) and his B.Sc. from Simon Fraser University (SFU), Canada (1996-2001). Before that, he studied at the Sofia High School of Mathematics (SMG).


Prof. Vechev’s work spans the broad intersection of artificial intelligence and programming languages, including both theoretical and system aspects. He has given a number of invited and keynote talks on these topics at various conferences and workshops (watch presentations here and here). Together with his group at ETH Zurich, he built a number of systems, widely used in academia and industry, including the creation of the first high-level quantum programming language Silq (watch presentation here). Vechev has also co-founded three deep tech start-ups based on groundbreaking research, all of which are positively impacting our world. First, he co-founded DeepCode which built a revolutionary AI system that automatically learned to detect security violations and errors in programs at a scale far beyond what was previously possible. DeepCode was initiated after Vechev’s group helped shape the research area of AI for code, part of ERC Starting Grant: AI for Programming Interview for the ERC. DeepCode was acquired in 2020 by the unicorn cybersecurity giant Snyk. In 2017 he co-founded ChainSecurity which was also acquired in 2020 by PwC and which was the first to bring formal trust and mathematical security guarantees to decentralized systems via advanced automated reasoning. Finally, in 2020, Martin Vechev co-founded LatticeFlow which creates new kinds of systems that enable the development of robust, reliable, and safe AI models, solving a pressing societal problem.

Prof. Vechev’s research has received numerous awards and recognitions including the ACM SIGPLAN Robin Milner Young Researcher Award for outstanding contributions in the area of programming languages, Distinguished Paper Awards (RV, OOPSLA), SIGPLAN Research Highlight, CACM Research Highlight, ERC Starting Grant in 2015 and ERC Consolidator Grant in 2022 (the only Bulgarian scientist with 2 ERC grants; interviews in Bulgarian here and here), Google and Facebook Faculty Research Awards, Outstanding Artifact Award (OOPSLA), IBM Research Division Award, IBM Extraordinary Accomplishment Award (awarded by John Kelly, IBM Senior Vice President and Director of IBM Research), IBM Research Outstanding Project Award, John Atanasoff Award, awarded by the President of Bulgaria (see photo, National TV news), IBM Invention and Outstanding Technical Achievement Awards, University Scholarships (SFU, Cambridge), and others.

Prof. Vechev was the Program Chair for ACM PLDI’17, a top conference in programming languages and systems, and the youngest program chair in the 40-year history of the conference. He has served on the SIGPLAN steering committee, SIGPLAN research highlights committee, and the SIGPLAN Doctoral Dissertation Award committee.


He has spoken on topics related to artificial intelligence, democratizing science, research and innovation at various venues, including the European Parliament and the launch of Horizon Europe (a 90+ billion euro program): interviews here and here.

2026

Ivo Petrov, Jasper Dekoninck, Dimitar Iliev Dimitrov, Martin Vechev
Not All Proofs Are Generated Equal: Evaluating Proof Quality Beyond Correctness
In: The Fortieth Annual Conference on Neural Information Processing Systems (NeurIPS) 2026, Dataset and Benchmarks Track

Ivo Petrov, Jasper Dekoninck, Martin Vechev
BrokenMath: A Benchmark for Sycophancy in Theorem Proving with LLMs
In: Forty-Third International Conference on Machine Learning (ICML 2026) (Spotlight)

 

Hanna Yukhymenko, Anton Alexandrov, Martin Vechev
Recovered in Translation: Efficient Pipeline for Automated Translation of Benchmarks and Datasets
In: The 64th Annual Meeting of the Association for Computational Linguistics (ACL 2026)

Jasper Dekoninck, Ivo Petrov, Kristian Minchev, Mislav Balunovic, Martin Vechev, Miroslav Marinov, Maria Drencheva, Lyuba Konova, Milen Shumanov, Kaloyan Tsvetkov, Nikolay Drenchev, Lazar Todorov, Kalina Nikolova, Nikolay Georgiev, Vanesa Kalinkova, Margulan Ismoldayev
The Open Proof Corpus: A Large-Scale Study of LLM-Generated Mathematical Proofs
In: The Fourteenth International Conference on Learning Representations (ICLR 2026)

2025

Maria Drencheva, Ivo Petrov, Maximilian Baader, Dimitar I. Dimitrov, Martin Vechev
GRAIN: Exact Graph Reconstruction from Gradients
In: International Conference on Learning Representations (ICLR 2025)

Mislav Balunović, Jasper Dekoninck, Nikola Jovanović, Ivo Petrov, Martin Vechev
MathConstruct: Challenging LLM Reasoning with Constructive Proofs
In: International Conference on Machine Learning (ICML 2025); ICLR 2025 (Workshop)

Ivo Petrov, Jasper Dekoninck, Lyuben Baltadzhiev, Maria Drencheva, Kristian Minchev, Mislav Balunović, Nikola Jovanović, Martin Vechev
Proof or Bluff? Evaluating LLMs on 2025 USA Math Olympiad
In: International Conference on Machine Learning (ICML 2025) (Workshop)

Jasper Dekoninck, Ivo Petrov, Kristian Minchev, Mislav Balunovic, Martin Vechev, Miroslav Marinov, Maria Drencheva, Lyuba Konova, Milen Shumanov, Kaloyan Tsvetkov, Nikolay Drenchev, Lazar Todorov, Kalina Nikolova, Nikolay Georgiev, Vanesa Kalinkova, Margulan Ismoldayev
The Open Proof Corpus: A Large-Scale Study of LLM-Generated Mathematical Proofs
In: International Conference on Machine Learning (ICML 2025) (Workshop)

Csaba Dékány, Stefan Balauca, Robin Staab, Dimitar I. Dimitrov, Martin Vechev
MixAT: Combining Continuous and Discrete Adversarial Training for LLMs
In: The Thirty-Ninth Annual Conference on Neural Information Processing Systems (NeurIPS 2025)

Mislav Balunović, Jasper Dekoninck, Ivo Petrov, Nikola Jovanović, Martin Vechev
MathArena: Evaluating LLMs on Uncontaminated Math Competitions
In: NeurIPS 2025 (Dataset and benchmark track)

Dimitar I. Dimitrov, Maximilian Baader, Mark Niklas Müller, Martin Vechev
SPEAR++: Scaling Gradient Inversion via Sparse Dictionary Learning
In: NeurIPS 2025 (Workshop)

Stefan Balauca, Mark Niklas Müller, Yuhao Mao, Maximilian Baader, Marc Fischer, Martin Vechev
Gaussian Loss Smoothing Enables Certified Training with Tight Convex Relaxations
In: Transactions on Machine Learning Research (TMLR 2025)

Yuhao Mao, Stefan Balauca, Martin Vechev
CTBENCH: A Library and Benchmark for Certified Training
In: 2025 Forty-Second International Conference on Machine Learning (ICML 2025)

2023

Florian E. Dorner, Nikola Konstantinov, Georgi Pashaliev, Martin Vechev
Incentivizing Honesty among Competitors in Collaborative Learning and Optimization
In: Conference on Neural Information Processing Systems (NeurIPS 2023)