AFTD Auto-Formalizing Theoretical Domains

op_1370_blocked_witness_minor_counterexample

theoremverified

For the two singleton families represented by masks 10 and 12 on four elements, mask 3 is incomparable to both. If one retains precisely the complement orbits for which both orientations are comparable to one of those two sets, then the principal matrix on masks 10, 12, and 3 has rows (1,0,1), (0,1,0), and (1,0,1), for any diagonal weights q. Thus the principal extension by this particular incomparable witness is singular.

Statement

theorem op_1370_blocked_witness_minor_counterexample (q : ℕ → ℚ) :
  (op1370_mask_inc 10 12 ∧ op1370_mask_inc 3 10 ∧ op1370_mask_inc 3 12) ∧
  (
    op1370_blocked_entry q 10 10 = 1 ∧ op1370_blocked_entry q 10 12 = 0 ∧
    op1370_blocked_entry q 10 3 = 1 ∧ op1370_blocked_entry q 12 10 = 0 ∧
    op1370_blocked_entry q 12 12 = 1 ∧ op1370_blocked_entry q 12 3 = 0 ∧
    op1370_blocked_entry q 3 10 = 1 ∧ op1370_blocked_entry q 3 12 = 0 ∧
    op1370_blocked_entry q 3 3 = 1)
Source
Counterexample to a fixed-witness blocked-matrix rank shortcut for arXiv:2611.12199, Conjecture 4.3. Masks encode subsets of Fin 4 in binary.
Verified
13 Nov 2026
Axioms
Classical.choiceQuot.soundpropext

The proof, explained

To show that the principal extension on masks 10, 12, and 3 is singular for any diagonal weights q, the proof first establishes that no blocked mask can simultaneously contain one index mask and be contained in another. Consequently, the diagonal weights q do not contribute to the matrix entries, reducing the calculation for arbitrary q to the constant weight q = 0. Finally, direct finite evaluation verifies that masks 10, 12, and 3 are pairwise incomparable and that the resulting matrix has identical first and third rows (1, 0, 1), rendering it singular.

  1. No 4-bit mask between mask i and mask j is blocked automation

    The claim is verified by exhaustive check over all 3 × 3 × 16 combinations using the kernel evaluator.

    line 30

          op1370_mask_sub c.val (op1370_mask_index j)) := by decide +kernel
  2. Each entry of the principal matrix is independent of the diagonal weights q

    Expanding the definition of the entry, each summand has c containing mask i. If c is also blocked, it cannot be contained in mask j by op1370_no_diagonal_term, so the q-term vanishes.

    lines 35–43

    uses op1370_no_diagonal_term

      unfold op1370_blocked_entry
      apply Finset.sum_congr rfl
      intro c _
      by_cases h : op1370_blocked c.val ∧ op1370_mask_sub (op1370_mask_index i) c.val
      · have no : ¬ op1370_mask_sub c.val (op1370_mask_index j) := by
          intro hc
          exact op1370_no_diagonal_term i j c ⟨h.1, h.2, hc⟩
        simp [h, no]
      · simp [h]
  3. The masks 10, 12, and 3 are pairwise incomparable automation

    This is verified by direct calculation using the kernel evaluator.

    line 44

    lemma op1370_blocked_witness_inc : op1370_mask_inc 10 12 ∧ op1370_mask_inc 3 10 ∧ op1370_mask_inc 3 12 := by decide +kernel
  4. The theorem reduces to checking the nine entries of the principal matrix

    We use independence from q to set q = 0 and evaluate the entries directly.

    lines 54–71

      refine ⟨op1370_blocked_witness_inc, ?_⟩
      have h00 := op1370_entry_irrelevant q 0 0
      have h01 := op1370_entry_irrelevant q 0 1
      have h02 := op1370_entry_irrelevant q 0 2
      have h10 := op1370_entry_irrelevant q 1 0
      have h11 := op1370_entry_irrelevant q 1 1
      have h12 := op1370_entry_irrelevant q 1 2
      have h20 := op1370_entry_irrelevant q 2 0
      have h21 := op1370_entry_irrelevant q 2 1
      have h22 := op1370_entry_irrelevant q 2 2
      simp only [op1370_mask_index, Matrix.cons_val_zero, Matrix.cons_val_one, Matrix.cons_val_two] at *
      change op1370_blocked_entry q 10 3 = op1370_blocked_entry (fun _ => 0) 10 3 at h02
      change op1370_blocked_entry q 12 3 = op1370_blocked_entry (fun _ => 0) 12 3 at h12
      change op1370_blocked_entry q 3 10 = op1370_blocked_entry (fun _ => 0) 3 10 at h20
      change op1370_blocked_entry q 3 12 = op1370_blocked_entry (fun _ => 0) 3 12 at h21
      change op1370_blocked_entry q 3 3 = op1370_blocked_entry (fun _ => 0) 3 3 at h22
      rw [h00,h01,h02,h10,h11,h12,h20,h21,h22]
      decide +kernel
op1370_no_diagonal_term
For any two indices i, j ∈ {0, 1, 2} and any 4-bit mask c, it is impossible for c to be blocked while containing mask i and being contained in mask j.

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

open Finsetset_option maxHeartbeats 2000000set_option maxRecDepth 8192abbrev op1370_mask_sub (a b : ℕ) : Prop := (a &&& b) = aabbrev op1370_mask_inc (a b : ℕ) : Prop := ¬ op1370_mask_sub a b ∧ ¬ op1370_mask_sub b aabbrev op1370_blocked (c : ℕ) : Prop :=  ¬ (op1370_mask_inc c 10 ∧ op1370_mask_inc c 12) ∧  ¬ (op1370_mask_inc (15 - c) 10 ∧ op1370_mask_inc (15 - c) 12)def op1370_mask_parity (a c : ℕ) : ℚ :=  if ((Finset.range 4).filter (fun i => Nat.testBit c i ≠ Nat.testBit a i)).card % 2 = 0 then 1 else -1def op1370_mask_weight (c : ℕ) : ℚ := if c < 8 then 2 else 1 / 2def op1370_blocked_entry (q : ℕ → ℚ) (a b : ℕ) : ℚ :=  ∑ c : Fin 16, if op1370_blocked c.val ∧ op1370_mask_sub a c.val then    op1370_mask_parity a c.val *      ((if op1370_mask_sub c.val b then q c.val else 0) +       (if op1370_mask_sub (15 - c.val) b then 1 else 0))  else 0def op1370_mask_index : Fin 3 → ℕ := ![10, 12, 3]lemma op1370_no_diagonal_term : ∀ i j : Fin 3, ∀ c : Fin 16,    ¬ (op1370_blocked c.val ∧ op1370_mask_sub (op1370_mask_index i) c.val ∧      op1370_mask_sub c.val (op1370_mask_index j)) := by decide +kernellemma op1370_entry_irrelevant (q : ℕ → ℚ) (i j : Fin 3) :    op1370_blocked_entry q (op1370_mask_index i) (op1370_mask_index j) =      op1370_blocked_entry (fun _ => 0) (op1370_mask_index i) (op1370_mask_index j) := by  unfold op1370_blocked_entry  apply Finset.sum_congr rfl  intro c _  by_cases h : op1370_blocked c.val ∧ op1370_mask_sub (op1370_mask_index i) c.val  · have no : ¬ op1370_mask_sub c.val (op1370_mask_index j) := by      intro hc      exact op1370_no_diagonal_term i j c ⟨h.1, h.2, hc⟩    simp [h, no]  · simp [h]lemma op1370_blocked_witness_inc : op1370_mask_inc 10 12 ∧ op1370_mask_inc 3 10 ∧ op1370_mask_inc 3 12 := by decide +kerneltheorem op_1370_blocked_witness_minor_counterexample (q : ℕ → ℚ) :  (op1370_mask_inc 10 12 ∧ op1370_mask_inc 3 10 ∧ op1370_mask_inc 3 12) ∧  (    op1370_blocked_entry q 10 10 = 1 ∧ op1370_blocked_entry q 10 12 = 0 ∧    op1370_blocked_entry q 10 3 = 1 ∧ op1370_blocked_entry q 12 10 = 0 ∧    op1370_blocked_entry q 12 12 = 1 ∧ op1370_blocked_entry q 12 3 = 0 ∧    op1370_blocked_entry q 3 10 = 1 ∧ op1370_blocked_entry q 3 12 = 0 ∧    op1370_blocked_entry q 3 3 = 1) := by  refine ⟨op1370_blocked_witness_inc, ?_⟩  have h00 := op1370_entry_irrelevant q 0 0  have h01 := op1370_entry_irrelevant q 0 1  have h02 := op1370_entry_irrelevant q 0 2  have h10 := op1370_entry_irrelevant q 1 0  have h11 := op1370_entry_irrelevant q 1 1  have h12 := op1370_entry_irrelevant q 1 2  have h20 := op1370_entry_irrelevant q 2 0  have h21 := op1370_entry_irrelevant q 2 1  have h22 := op1370_entry_irrelevant q 2 2  simp only [op1370_mask_index, Matrix.cons_val_zero, Matrix.cons_val_one, Matrix.cons_val_two] at *  change op1370_blocked_entry q 10 3 = op1370_blocked_entry (fun _ => 0) 10 3 at h02  change op1370_blocked_entry q 12 3 = op1370_blocked_entry (fun _ => 0) 12 3 at h12  change op1370_blocked_entry q 3 10 = op1370_blocked_entry (fun _ => 0) 3 10 at h20  change op1370_blocked_entry q 3 12 = op1370_blocked_entry (fun _ => 0) 3 12 at h21  change op1370_blocked_entry q 3 3 = op1370_blocked_entry (fun _ => 0) 3 3 at h22  rw [h00,h01,h02,h10,h11,h12,h20,h21,h22]  decide +kernel
See it in the graphDiscuss this resultChallenge it