Etienne Bamas

Hello!

Welcome to my webpage. I am a post-doctoral researcher at EPFL, affiliated to the group of Prof. Emmanuel Abbé, and a contractor at Apple MLR. I am working in AI reasoning and AI for mathematics. My latest projects are about improving post-training techniques for formal theorem proving in Lean with LLMs, either to improve theorem proving performance or to speed-up the actual training.

Previously, I used to work in Theoretical Computer Science. You can find more info about me here.

Contact

• Email: firstname dot lastname at epfl dot ch

Publications

Most of my publications can also be found on my scholar profile.

• L. Milikic, E. Bamas, V. Kunčak, E. Abbe. Learning Lean Feedback for Efficient Proof-Candidate Evaluation on GPUs.
  [PDF], VeriCodeGen workshop@NeurIPS 2026.

• L. Milikic, E. Bamas, V. Shashkov, R.-D. Buhai, R. Vintan, O. Svensson, E. Abbe, V. Kunčak. Formal Theorem Proving with Lean World Models.
  [PDF], under review.

• M. Santelmo, X. Wei, I. Fakih, F. Bauer, J. G. Giraldo, C. Li, E. Bamas, E. Abbe. Blind-Spots-Bench: Evaluating Blind Spots in Multimodal Models.
  [arXiv].

• K. Brilliantov, E. Bamas, E. Abbe. k-server-bench: Automating Potential Discovery for the k-Server Conjecture.
  [arXiv], 3rd AI for Math Workshop@ICML 2026.

• E. Bamas, S. Li, L. Rohwedder. Randomized Rounding over Dynamic Programs.
  [arXiv], in STOC 2026.

• E. Bamas. Lift-and-Project Integrality Gaps for Santa Claus.
  [arXiv], in SODA 2025.

• E. Bamas, S. Morell, L. Rohwedder. The Submodular Santa Claus Problem.
  [arXiv], in SODA 2025.

• E. Bamas, SG Nagarajan, O. Svensson. An Analysis of $D^alpha$ seeding for k-means.
  [arXiv, conference version], in ICML 2024.

• E. Bamas, A. Lindermayr, N. Megow, L. Rohwedder, J. Schloeter. Santa Claus meets Makespan and Matroids: Algorithms and Reductions.
  [arXiv], in SODA 2024 (invited to TALG special issue).

• E. Bamas, L. Rohwedder. Better Trees for Santa Claus.
  [arXiv], in STOC 2023.

• E. Bamas, M. Drygala, O. Svensson. A Simple LP-Based Approximation Algorithm for the Matching Augmentation Problem.
  [arXiv], in IPCO 2022.

• E. Bamas, M. Drygala, A. Maggiori. An Improved Analysis of Greedy for Online Steiner Forest.
  [arXiv], in SODA 2022 (student authors only).

• E. Bamas, P. Garg, L. Rohwedder. The Submodular Santa Claus Problem in the Restricted Assignment Case.
  [arXiv], in ICALP 2021.

• E. Bamas, A. Maggiori, O. Svensson. The Primal-Dual method for Learning Augmented Algorithms.
  [arXiv], in NeurIPS 2020 (oral presentation, top 1% of submissions).

• E. Bamas, A. Maggiori, L. Rohwedder, O. Svensson. Learning Augmented Energy Minimization via Speed Scaling.
  [arXiv], in NeurIPS 2020 (spotlight presentation, top 3% of submissions).

• E. Bamas, L. Esperet. Local Approximation of the Maximum Cut in Regular Graphs.
  [arXiv], in WG 2019, journal version in Theoretical Computer Science.

• E. Bamas, L. Esperet. Distributed coloring of graphs with an optimal number of colors.
  [arXiv], in STACS 2019.

Projects and advising

I also advised several student projects/MSc theses either on algorithms and theory or on more applied projects (see poster for a more applied example)