Skip to content

Non-dense fitting B: every E8 term with x·y ≠ 0 has rank ≥ 9; square root-code floors E8 ≥ 72, P112 ≥ 64, P104 ≥ 56, D4 ≥ 12 for any fitting B (no κ claim) - #3045

Open
DreamingOfClouds wants to merge 1 commit into
CrocSwap:mainfrom
DreamingOfClouds:research/nondense-pairing-square-floors
Open

DreamingOfClouds wants to merge 1 commit into
CrocSwap:mainfrom
DreamingOfClouds:research/nondense-pairing-square-floors

Conversation

@DreamingOfClouds

Copy link
Copy Markdown
Contributor

What this adds

Only research/nondense-pairing-square-floors/. Extends the dense-pairing floor of research/ideal-decoder-floors/ (Theorem 4) to any fitting B (B_tt = 1, B_ts = 0 on non-orthogonal pairs, free on orthogonal pairs) for square root-code decoders (m = T, XYᵀ = I). Related: #968, #1744 (rank ≥ 9 on P111/P112 in its scope), #2176, #2338 (Theorems P, R).

  • Lemmas 6.1-6.3 (PROVEN, written proofs): per-term fitting minrank = mr₂ of the orthogonality graph on R∩A; normal-pair reduction; orthogonal-set deletion lowers mr₂ by ≤ 1.
  • Theorem 6 (CHECKED): all 4022 non-orthogonal root-hyperplane pair types of E8 (440 880 hyperplanes, W_J = Stab(n₁)-orbits, min |D| = 30) are UNSAT at fitting rank ≤ 8, each with a drat-trim VERIFIED proof; two independent orbit implementations and two independent CNF writers; brute force over all fitting B agrees with both writers on 60 instances.
  • Corollary 6.4: square root-code decoders cost ≥ 72 (E8), ≥ 64 (P112), ≥ 56 (P104), ≥ 12 (D4) for any fitting B, all sharp (exact decoders re-built and checked).
  • Counterexamples to the per-term claim (CHECKED): E8 orthogonal normals |D| = 24, mr₂ = 6; D6 non-orthogonal types with mr₂ 3-4 (scan |D| ≤ 19 only); one E7 type |D| = 11, mr₂ = 5. No decoder below any floor is shown (B is shared across terms).
  • Open / not claimed: overcomplete decoders with x_j ⊥ y_j terms, E7/D6 square (global-B coupling), factors outside the root code. No κ claim.

Verification

bash verify.sh (a few minutes, ≤ 2 processes): manifest, vendored drat-trim build, both orbit recounts and their agreement, fixed 32-type sample re-solved by both writers with DRAT VERIFIED, brute-force agreement, the counterexamples, the sharp decoders. FULL=1 bash verify.sh re-solves all 4022 types with both writers, every proof drat-trim VERIFIED; it was run in full at this head and passes (log in evidence/verify_FULL.log). PROOF_MAP.md lists claim → check; anything unchecked is CONDITIONAL in the README.

Prepared by DreamingOfClouds with Claude Opus 5.5 (Anthropic), via Claude Code; Apache-2.0.

🤖 Generated with Claude Code

…ot-code floors E8 72, P112 64, P104 56, D4 12 (no kappa claim)

Lemmas 6.1-6.3 with proofs, Theorem 6 (4022 types, two orbit implementations, two CNF writers, DRAT), sharp decoders, per-term counterexamples (E8 orthogonal, D6, E7).

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant