Mislav Balunović

I am a researcher working on AI for mathematics and reasoning with a focus on improving theorem proving capabilities of LLMs. I’ve created MathArena (matharena.ai), a benchmark which evaluates LLMs on latest math competitions which is now used by leading labs such as Google DeepMind, xAI and Microsoft. Besides my research, I have a gold medal from IMO and silver medals from IOI.

Notes3

• Joined Google DeepMind as Senior Research Scientist in September 2025, moving to industry after his PhD and postdoc at ETH Zurich's SRI lab.Sep 1, 2025
• Created MathArena (matharena.ai), a benchmark evaluating LLMs on recent uncontaminated math competitions, now reportedly used by Google DeepMind, xAI and Microsoft - his most-cited public artifact.Feb 7, 2025
• AgentDojo, a benchmark for AI agent safety, won first prize in the ML Safety SafeBench competition (USD 50,000) in May 2025.May 7, 2025
• Won a gold medal at the International Mathematical Olympiad (Croatia) and silver medals at the International Olympiad in Informatics during high school; the IMO background directly seeded his later AI-for-mathematics research.2013-2014
• Co-founded Invariant Labs (AI agent security) and served as CTO; his ongoing agent-safety work (AgentDojo) reflects that startup's focus.
• PhD at ETH Zurich (Apr 2019 - Apr 2024) in the Secure, Reliable and Intelligent Systems Lab supervised by Martin Vechev; thesis on trustworthy and reliable machine learning.Apr 2024
• His research on LLM privacy ("Language Models are Advanced Anonymizers") and reasoning benchmarks has been covered by Ars Technica, Forbes and Wired.2025
• Received the ETH Medal for Outstanding Master Thesis.2019
• Held engineering internships at Twitter, Facebook and SigOpt, giving him an industry track alongside academic research.