TMIP* = RE, and the Aldous–Lyons conjecture is false #
The main theorem of The Aldous–Lyons Conjecture II: Undecidability (Bowen–Chapman–Vidick,
arXiv:2501.00173, thm:tailored_MIP*=RE, II:1505), proved:
TailoredGameValue.tailored_halting_reduction is the Mathlib-only statement
TailoredGameValue.TailoredHaltingReduction of MIPRE/TailoredGameValue.lean (blueprint
thm:tmip-re). Tailored compression is built from question reduction
(Intro.Inhabit.tailoredIntrospection), answer reduction
(AnsRed.Typed.tailoredAnswerReduction) and parallel repetition (tailoredRepetition), and the
class theorems follow for both readings of TMIP*.
With paper I (The Aldous–Lyons Conjecture I: Subgroup Tests, Bowen–Chapman–Lubotzky–Vidick,
arXiv:2408.00110), the sofic value of subgroup tests is not approximable and the Aldous–Lyons
conjecture is false: SubgroupTestValue.aldous_lyons_false is the negation of the Mathlib-only
statement SubgroupTestValue.AldousLyons of MIPRE/SubgroupTestValue.lean (blueprint
thm:aldous-lyons-false).
Tailored gap compression (blueprint thm:tailored-compression-target), inhabited.
Equations
Instances For
TMIP* = RE (paper II, thm:tailored_MIP*=RE, II:1505; blueprint thm:tmip-re): the
halting problem reduces to tailored games, with a perfect ZPC strategy on the machines that halt
and synchronous value at most 1/2 on the others.
The halting problem reduces to tailored games, in the quantum value.
TMIP* = RE, computable class (II:787, II:6997).
TMIP* = RE, the paper's polynomial-time class (II:787).
The sofic value of subgroup tests is not approximable (paper I, Corollary I:2144, measure-free).
The Aldous–Lyons conjecture is false (paper I, Corollary I:2144, with paper II's main
theorem; blueprint thm:aldous-lyons-false).