repair_drift_separable_identity
For all real numbers f, H, and ς with f ≠ 0, H ≠ 0, and f + H ≠ 0, the sum of the memory gradient ((descent_threshold f H - ς) / H²) multiplied by the memory repair drift (f - H) and the satisfaction gradient (1 / H) multiplied by the satisfaction repair drift (1 + H - ς) equals (f / H²) * (scalar_repair_certificate f H - ς).
Statement
theorem repair_drift_separable_identity (f H ς : ℝ) (hf : f ≠ 0) (hH : H ≠ 0) (hfH : f + H ≠ 0) :
((descent_threshold f H - ς) / H^2) * (f - H) + (1 / H) * (1 + H - ς) =
(f / H^2) * (scalar_repair_certificate f H - ς)- Formalizes
- arXiv:2611.06076:974
- Verified
- 08 Nov 2026
- Axioms
Classical.choiceQuot.soundpropext- Built on 2
- descent_thresholdThe continuous-time descent satisfaction threshold S_f(H) for activation budget f and memory weight H is (1 + H)² / (f + H).
- scalar_repair_certificateThe scalar repair progress certificate K_f(H) for activation budget f and memory weight H is (1 + H) * (f + (2f - 1)H) / (f(f + H)).
Read back from the Lean
For all real numbers $f$, $H$, and $ς$, if $f \neq 0$, $H \neq 0$, and $f + H \neq 0$, then $$\left(\frac{\text{descent\_threshold}(f, H) - ς}{H^2}\right)(f - H) + \frac{1}{H}(1 + H - ς) = \frac{f}{H^2}(\text{scalar\_repair\_certificate}(f, H) - ς).$$
Written by a model that saw only the Lean, never the English above. If the two disagree, that is worth a challenge.
The proof, explained
The theorem establishes an algebraic identity between rational functions of f, H, and ς that governs the drift of separable potential functions along repair dynamics. The proof unfolds the definitions of descent_threshold and scalar_repair_certificate to express the equation as an identity between rational expressions. Using the hypotheses that f, H, and f + H are non-zero, denominators are cleared, reducing the problem to a polynomial identity that is verified by expanding both sides.
The equation is stated explicitly in terms of elementary rational functions of f, H, and ς.
Unfolding the definitions of descent_threshold as (1 + H)² / (f + H) and scalar_repair_certificate as (1 + H) * (f + (2 * f - 1) * H) / (f * (f + H)) replaces the high-level potential drift concepts with explicit rational expressions.
unfold descent_threshold scalar_repair_certificateClearing denominators reduces the equality to a polynomial identity, which holds upon expanding both sides. automation
The `field_simp` tactic uses the hypotheses f ≠ 0 (hf), H ≠ 0 (hH), and f + H ≠ 0 (hfH) to clear all denominators across the equation, reducing it to the polynomial equality ((1 + H)² - (f + H) * ς) * (f - H) + H * (f + H) * (1 + H - ς) = (1 + H) * (f + H * (2 * f - 1)) - f * (f + H) * ς. The `ring` tactic then closes the goal by expanding both sides and checking that their coefficients coincide.
field_simp ring
- The clearing of denominators and verification of the polynomial identity are handled entirely by automation (`field_simp` and `ring`), which rely on the non-vanishing hypotheses f ≠ 0, H ≠ 0, and f + H ≠ 0.
Written by a model (gemini-3.8-flash) from the Lean proof and the goal states Lean reported at each step. Lean verified the proof; this explanation is a reading aid and is not itself machine-checked.
Lean source view module on GitHub
Choose a tactic — a marked line — to see the goals Lean had before and after it, as in Lean’s infoview. Step with ← → or j k.
/-- Drift identity for separable potential functions along repair dynamics. -/theorem repair_drift_separable_identity (f H ς : ℝ) (hf : f ≠ 0) (hH : H ≠ 0) (hfH : f + H ≠ 0) : ((descent_threshold f H - ς) / H^2) * (f - H) + (1 / H) * (1 + H - ς) = (f / H^2) * (scalar_repair_certificate f H - ς) := by unfold descent_threshold scalar_repair_certificate field_simp ring