Get Started
Home
Topics
Search
Library
Research questionHow can uncoupled learning keep individual regret bounded independently of time in arbitrary multi-player normal-form games?In arbitrary multi-player normal-form games, individual regret under uncoupled learning may continue to grow as play proceeds. Obtaining a horizon-independent guarantee is difficult when strategic interactions involve several players.
Machine Learning
Multi-agent Systems
Reinforcement Learning
Research Paper
Latest papersRecent research connected to this question, newest first.Constant regret in general games via higher-order optimismThe setting includes N-player normal-form games with up to K actions per player, where all players use the uncoupled learning procedure. The reported theoretical guarantee is O(N^3 log^2 K) individual regret uniformly over the horizon; the source describes a higher-order optimistic FTRL construction but supplies no evidence beyond this setting.research paper · Sep 3, 2026
Related questions
How can long-horizon LLM agents learn when to group actions without overcommitting?How can we learn welfare-optimal approximate Nash equilibria in concurrent stochastic games under uncertain transitions while certifying exact-equilibrium nonexistence?How can offline goal-conditioned reinforcement learning learn reliable values for long-horizon tasks without compounding overestimation?How can teams estimate whether multiclass accuracy gains reduce optimization regret before building a model?