PeopleofSuperintelligence

Researcher

Côme Fiegel

Côme Fiegel is an AI researcher whose work connects optimization, game theory, and language-model alignment. He completed doctoral research at ENS Ulm with Pierre Ménard and Vianney Perchet during 2022–2025, and holds a PhD from ENSAE, Institut Polytechnique de Paris. His 2023 game-tree research introduced Balanced FTRL and Adaptive FTRL and received ICML's Outstanding Paper award; later work applies mirror descent to extensive-form games, Nash learning to language-model feedback, and bandit algorithms to zero-sum games. A 2026 paper studies log-barrier methods for minimax policies, and as of October 2026 he is a postdoctoral researcher at ENSTA with E. Berthier, working on optimization, bandits, and games.

Now at
ENSTA Paris

Timeline

  1. 2020

    Kometo for multi-fidelity optimization

    In 2020, he coauthors Adaptive multi-fidelity optimization with fast learning rates and presents Kometo. The algorithm matches the rates up to logarithmic factors without prior knowledge of function smoothness or fidelity assumptions.

  2. 2022–2025

    Doctoral research at ENS Ulm

    From 2022 to 2025, he completes doctoral research at ENS Ulm with Pierre Ménard and Vianney Perchet.

  3. 2023

    Balanced FTRL and an ICML award

    In 2023, he coauthors Adapting to game trees in zero-sum imperfect information games, which proposes Balanced FTRL and Adaptive FTRL for learning through self-play with trajectory feedback. ICML gives the paper its Outstanding Paper award.

  4. 2024

    Mirror descent for extensive-form games

    In 2024, he coauthors Local and Adaptive Mirror Descents in Extensive-Form Games. The paper applies adaptive online mirror descent at each information set, uses individually decreasing learning rates and a regularized loss, and uses fixed sampling to reduce variance under trajectory feedback.

  5. 2024

    Nash learning from human feedback

    In 2024, he coauthors Nash Learning from Human Feedback, which learns a pairwise preference model and trains language-model policies toward a Nash equilibrium. The authors test the approach on text summarization.

  6. 2025

    Bandit feedback in zero-sum games

    In 2025, he coauthors The Harder Path, which proposes two uncoupled algorithms for last-iterate convergence to a Nash equilibrium in zero-sum matrix games with bandit feedback.

  7. 2026

    Log-barrier methods for minimax policies

    In 2026, he coauthors a paper that uses log-barrier regularization and dual-focused analysis to reach high-probability convergence for minimax policies in zero-sum matrix games.

  8. As of October 2026

    Postdoctoral researcher at ENSTA

    As of October 2026, he is a postdoctoral researcher at ENSTA, where he works with E. Berthier on optimisation, bandits, and games. He holds a PhD from ENSAE, Institut Polytechnique de Paris.

Career

  1. Since 2026ENSTA ParisPostdoctoral researcher

Education

  1. 2025École Normale SupérieurePhD
  2. Institut Polytechnique de ParisPhD

From the biography.

  1. The Massachusetts Institute of Technology trained or employed 44 of the 187 AI major-prize winners, more than any other universityIncludes Côme Fiegel. The Massachusetts Institute of Technology trained or employed 44 of the 187 AI major-prize winners, more than any other university. We compare doctorate and 5+ year employment links, career timing and current workplaces for 187 major-prize and 1,111 best-paper winners.
  2. More people shaping AI studied at Massachusetts Institute of Technology than at any other institutionIncludes Institut Polytechnique de Paris. We estimate where the people shaping AI studied and where their careers led. Massachusetts Institute of Technology leads with 467 former students, including 291 recorded doctorates.
  3. The company share of AI researchers’ top papers was 31% in 2020–2024 and 8% in 1970–1974Includes Institut Polytechnique de Paris. We estimate where the people shaping AI worked when their most-cited papers appeared: the company share was 31% in 2020–2024 and 8% in 1970–1974, and Google led all organizations in 2020–2024.
  4. The median h-index is 48 at universities and 14 at frontier AI labsIncludes Institut Polytechnique de Paris. The median h-index is 48 at universities and 14 at frontier AI labs. We compare publishing careers, recent citations, dual appointments and job moves across 2,420 university researchers and 308 frontier-lab researchers.

Colleagues

People named in this biography, and people with dated positions at the same organization in the same years.

Sources and credits

Photo: We have no published photo of Côme Fiegel yet. Send us a free-licensed photo.

  1. Adaptive multi-fidelity optimization with fast learning rates proceedings.mlr.press. Checked 22 October 2026.
  2. Michal Valko | ELLIS Fellow | AI Researcher, RL and LLMs misovalko.github.io. Checked 22 October 2026.
  3. Adapting to game trees in zero-sum imperfect information games proceedings.mlr.press. Checked 22 October 2026.
  4. ICML 2023 Awards icml.cc. Checked 22 October 2026.
  5. Local and Adaptive Mirror Descents in Extensive-Form Games proceedings.neurips.cc. Checked 22 October 2026.
  6. Nash Learning from Human Feedback proceedings.mlr.press. Checked 22 October 2026.
  7. The Harder Path: Last Iterate Convergence for Uncoupled Learning in Zero-Sum Games with Bandit Feedback proceedings.mlr.press. Checked 22 October 2026.
  8. Optimal last-iterate convergence in matrix games with bandit feedback using the log-barrier arxiv.org. Checked 22 October 2026.
  9. Projects and Team perso.ensta.fr. Checked 22 October 2026.