Skip to content

Refined recovery: max fiber = zero fiber under an ε₀ threshold; exact depth-27 optimum (1+12ρ₂²⁵+3ρ₂²⁶)·F²⁷, 696 bits (no κ) - #6848

Merged
Mister-Meeseeks merged 1 commit into
CrocSwap:mainfrom
apoursh:research/refined-recovery-max-fiber
Nov 10, 2026
Merged

Mister-Meeseeks merged 1 commit into
CrocSwap:mainfrom
apoursh:research/refined-recovery-max-fiber

Conversation

@apoursh

@apoursh apoursh commented Nov 7, 2026

Copy link
Copy Markdown
Contributor

Scoped proof with exact finite corroboration; no κ. Package: research/refined-recovery-max-fiber.

Result

Theorem. Take any depth and any frame sequence in G. Suppose its zero fiber is within ε* = (1/13)/((23/12)⁸−1) ≈ 4.25·10⁻⁴ (relative) of its Fourier main term 2^(dim D−4t)·13^(8t). Then max_y |fiber(y)| = |fiber(0)|.

Corollary (exact depth-27 optimum).

min_U max_y |fiber(y)| = (1 + 12ρ₂²⁵ + 3ρ₂²⁶)·F²⁷ ≈ 15.7897746·F²⁷  →  696 tag bits

This is the first exact value past PR6660's range t ≤ 26, and it answers PR6660's open question 1 at t = 27.

Why the box lemma is not enough, and what replaces it

For y ≠ 0 the per-stage boxes can depend on the chain history, and the overlapping stage cosets block PR6660's stage-by-stage count. The proof has four steps:

  1. Factorization. Poisson summation over H = ker Φ writes the fiber as a sum over h ∈ H^⊥. Each term is a product of eight one-coordinate signed chain sums Z_k, because each coordinate's box depends only on its own history.
  2. Ties are the only adaptivity. Without a rounding tie, each box is a fixed 13-interval shifted by an even amount, so |Z_k| equals its y = 0 value.
    • At ties, s = 4x+2. The state follows x ↦ −13x + 14r.
    • Only the 16 states x ∈ [−8,7] can still meet a 12-element box. All other states alternate between two classes forever, and their boxes stay fixed.
  3. Bounding the tie states. For those states Z is a product of 10×10 transfer matrices. An interval family on the 1/12 grid is exactly closed under T₀/13 and T₁, which gives |Z_k| ≤ (23/12)·13^(zeros) at every depth.
    • The constant is essentially attained (above 1.9166 at depth 27).
    • The naive majorant with constant 1 fails, e.g. Z₋₂₂(1,0,1) = 24.
  4. The penalty. Every state whose ratio exceeds 1 starts with a 12-element box, so the main term loses a factor 12/13. That loss beats (23/12)⁸·ε₀ whenever ε₀ ≤ ε*.

Verification

verify.py (stdlib, exact rationals) prints PASS_REFINED_RECOVERY_MAX_FIBER in about 45 s, in both normal and -O mode. It checks:

  • C and ρ₂, derived from the pinned M;
  • the box lemma for |s| ≤ 3000;
  • the transfer matrices against direct chain recursion (|s| ≤ 150, depth ≤ 4);
  • the closed interval family and the cap for each state;
  • the Poisson identity on six random t = 2 instances, with random D and tie coordinates;
  • ε* and ε₀ exactly, and the 696-bit count;
  • two controls: the failing naive majorant, and t = 26 direct sums (ε₀ ≈ 0.0128) lying outside the threshold. PR6660's box lemma covers those instead.

Both scripts/research_package.py --package and scripts/check_research_commit.py --check-source-history --check-modes pass.

Limitations

Credit

Nick Petalas (PR6660: chain bijection, depth law, the open question); CubeJerry (PR6327: chain model, F, box facts); zxv (#5444: M).

AI disclosure: Anthropic Claude (Claude Code) found the argument, wrote the verifier and ran the checks, under apoursh's direction.

🤖 Generated with Claude Code

… (no κ)

For every frame sequence whose zero fiber is within eps* ~ 4.25e-4 of its
Fourier main term, no output has a larger fiber than 0. Rounding ties are the
only adaptivity; a closed transfer-matrix interval family bounds tie
coordinates by 23/12, paid for by their 12-element first box. With CrocSwap#6828's
FRAMES27 (eps0 = 5.45e-6) the depth-27 optimum is exactly
(1+12 rho2^25+3 rho2^26) F^27, 696 bits.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
@Mister-Meeseeks
Mister-Meeseeks merged commit 2fc0d3b into CrocSwap:main Nov 10, 2026
2 checks passed
Mister-Meeseeks pushed a commit that referenced this pull request Nov 10, 2026
…-6 (no κ)

The main Fourier term is a universal lower bound on the zero fiber (positive
characters, rank <= 104). FRAMES27 of #6828 extended by U_j = I keeps rank 104
and cannot raise eps0, so #6848 gives a matching upper bound within 1+theta,
theta = 5.45e-6. Optimal tag = ceil(8t log2 13) - 104 bits (696, 725, 755,
785, ...), first possibly ambiguous depth 584977; 100 bits below identity
frames for 27<=t<=400.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Mister-Meeseeks pushed a commit that referenced this pull request Nov 10, 2026
…age (no κ)

OPT(t) = 2^-104 13^(8t) (1+eps_t). Exact first moment of S over PR6660's
I+V E frame family (psi_j uniform into the 100-dim complement R') gives
eps_t <= (2^104-16) q^t + 14*13^(-4t) + 13^(-8t) for t>=29 (below the #6848
threshold); Jensen on the uniform weight distribution of chi o phi_j gives
eps_t >= 2^104 13^(-4t) for all frames.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Mister-Meeseeks pushed a commit that referenced this pull request Nov 10, 2026
Integer I+VE frame sequences found by local search and certified exactly:
rank 104, full enumeration of D, zero-fiber excess S_t = 2.72e-7, 5.01e-9,
5.05e-12 over 13^(8t)/2^104, all below the #6848 threshold. Tightens #6850's
t=28 window 20-fold and replaces #6851's existential bounds at t=29,30.

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.

2 participants