Parallel Repetition of Entangled Games with Exponential Decay via the Superposed Information Cost
summary
Shows that for games with independent uniform questions, if one-shot entangled value equals 1 − ε then the n-fold entangled value falls exponentially with n, with rate roughly Ω(ε^2/(log K + log L)), so ω(G^n) ≤ exp(−Ω(n ε^2/(log K+log L))). Proof introduces the superposed information cost (SIC), proves SIC ≥ Ω(ε), uses mutual information additivity and a chain rule, and applies Uhlmann theorem* on purifications to show SIC zero implies value one, yielding the exponential decay.
AI-written summary · about
chapters (11)
- 0:00Entangled games introduction
- 3:11Parallel repetition entangled games
- 7:27Extra resources in entangled games
- 10:17Limitations of information cost
- 13:28Superposed information cost
- 16:41Superposed cost properties
- 21:43Superposed cost versus game value
- 25:22Proof insight with superposed states
- 28:58Winning strategies from superposed states
- 30:04Recap and conclusions
- 32:11Q&A
