← All works
Research paper

Mastering Atari, Go, Chess and Shogi by Planning with a Learned Model

At a glance

MuZero learns a model of a game to plan moves without being given its rules.

In one paragraph

Introduced MuZero, which learns a model of a game from experience and uses it to plan moves, without being given the game rules. It matched AlphaZero in chess, shogi and Go, and set a new benchmark across 57 Atari games.

Authors 12

  1. Julian Schrittwieser
  2. Ioannis Antonoglou
  3. Thomas Hubert
  4. Karén Simonyan
  5. Laurent Sifre
  6. Simon Schmitt
  7. Arthur Guez
  8. Edward Lockhart
  9. Demis Hassabis
  10. Thore Graepel
  11. Timothy Lillicrap
  12. David Silver

Organizations

Credited for this work at publication.