TY  - RPRT
TI  - Optimal last-iterate convergence in matrix games with bandit feedback using the log-barrier
AU  - Come Fiegel
AU  - Pierre Menard
AU  - Tadashi Kozuno
AU  - Michal Valko
AU  - Vianney Perchet
PY  - 2026
UR  - https://arxiv.org/abs/2604.15242
ID  - 2604.15242
ER  -