Reproduce “The Complexity of Strategizing Against No-Regret Learners: Pure When It Is Easy, Hard Even When Mixed” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Yiheng Su, Emmanouil-Vasileios Vlatakis-Gkaragkounis
The paper shows that planning against an online learner can remain computationally hard even when optimizer actions are mixed. In constructed general-sum games, it reduces fractional action counts to an independent set, proving no polynomial-time optimizer achieves cT−o(T) additive accuracy against the best pure sequence unless P=NP. In zero-sum games, pure responses retain regret-based guarantees under stated conditions.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.