Proof of Theorem 4, p.4 — the increment mean is decreasing,
ProvedPhysReaudit.increment_meanp2o-gran-per-chapterp2o-plan-paperp2o-v1physical-auditingsequential-testingstatistics
Let , , and . The increment mean
is strictly decreasing in , satisfies for every , and
So the SPRT statistic drifts upward at rate at least on the whole good class , which drives the bound on .
Formalization Note is stated on all of ; kl is the published binary relative entropy klBer.
Differences from the source:
- encoding: q₀ = (b − ε)/(κp), t = b/(κp) with 0 < ε < b < κp → any reals 0 < q₀ < t < 1; reason: exactly the pairs that arise (κ = a = 1, b = t, ε = t − q₀ realise any such pair), so the same instances
Preamble
import Mathlib import Definitions.Def_LimitedPriceChanges_LowerBound_KL import Definitions.Def_PhysReaudit_Delta open LimitedPriceChanges.LowerBound (klBer)
Formal statement
namespace PhysReaudit
/-- **The increment mean** (Q. Li, *When Can Physical Evidence Survive a World-Model Update?*, review revision of 3 November 2026 (sha256 a20c437b…), proof of Theorem 4, p.4 (PDF p.5)): "The increment mean
m(q) = qz− + (1 − q)z+ is decreasing in q and satisfies m(q) ≥ m(q0) = D > 0 on the good class."
For `0 < q₀ < t < 1`, the map `m(q) = q z₋ + (1 − q) z₊` is strictly decreasing on `ℝ`, `m(q₀) = kl(q₀, t)`,
and `kl(q₀, t) > 0`; hence `m(q) ≥ m(q₀) = D > 0` for every `q ≤ q₀`. -/
theorem increment_mean {q₀ t : ℝ} (h0 : 0 < q₀) (h1 : q₀ < t) (h2 : t < 1) :
StrictAnti (fun q : ℝ => q * zMinus q₀ t + (1 - q) * zPlus q₀ t) ∧
(∀ q ≤ q₀, q₀ * zMinus q₀ t + (1 - q₀) * zPlus q₀ t ≤ q * zMinus q₀ t + (1 - q) * zPlus q₀ t) ∧
q₀ * zMinus q₀ t + (1 - q₀) * zPlus q₀ t = klBer q₀ t ∧
0 < klBer q₀ t := by sorry
end PhysReaudit
Source
Q. Li, When Can Physical Evidence Survive a World-Model Update? Finite-cost re-auditing and reusable policy certificates, review revision of 3 November 2026 (STOC 2027 submission, sha256 a20c437b34115b84982339e9a6fa319cb3a66f66d9d4c474465cbb954c79093d), proof of Theorem 4, p.4 (PDF p.5), "The increment mean m(q) = qz− + (1 − q)z+ is decreasing in q and satisfies m(q) ≥ m(q0) = D > 0 on the good class."
Human review
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.