Appendix
–Neural Information Processing Systems
Our goal in this section is twofold. To avoid excessive repetition of notation and proof arguments, we purposefully leavethis sectionnot self-contained and only outline these adjustments needed. We refer an interested reader to the work of Auer et al.[18] for the full and detailedproof. Instead of aiming to find a black-box reduction to the classical problem, we therefore investigate in detail the proof of theΩ KTlower bound for the classical MAB due to Auer et al.[18, 14 Theorem 5.1] and observe that their argument goes through for our MA-MAB problem as well. Instead of repeating their proof, we survey the key steps of their proof in which they assume the algorithm tobedeterministically pulling anarmandhighlight whytheargument holdsevenwhen thisisnotthecase.
Neural Information Processing Systems
Feb-11-2026, 03:59:41 GMT