Skip to content

Add an exponent-bracket check with no logarithms, applied to the printed Ising, percolation and Sabra series - #58

Draft
orda-aska wants to merge 5 commits into
mainfrom
scaling-check-no-logs
Draft

orda-aska wants to merge 5 commits into
mainfrom
scaling-check-no-logs

Conversation

@orda-aska

@orda-aska orda-aska commented Nov 20, 2026 •

Copy link
Copy Markdown
Contributor

Fixes #57

What is added

Five new files. No existing file is edited, so this touches nothing an open pull request touches.

  • scaling-check/check.mjs: cmpPower (the sign of (y_b/y_a)^q - (x_b/x_a)^p in BigInt), exponentIn (a closed bracket in three modes), exponentInLog2 (the same for a printed column of base-2 logarithms), tightestHalfWidth, and an 80-digit logarithm used only by the tests. No dependencies.
  • scaling-check/series.mjs: the five series, each with the script that prints it, the commit (48b923b), the row and the column.
  • scaling-check/table.mjs: prints the verdicts. With verify it runs critical/table.mjs and turbulence/table.mjs quick and finds every pinned number again.
  • scaling-check/README.md: what the check certifies and what it does not, and the verdict on each series, including the ones it rejects.
  • tests/scaling-check.mjs: picked up by tests/run.mjs with no change to it.

Commands and output

  • node tests/scaling-check.mjs prints 40 lines marked PASS and ends with all scaling-check checks pass.
  • npm test ends with all 9 check file(s) pass. It takes about 45 seconds.
  • node scaling-check/table.mjs verify ends with every pinned number is in the output of the scripts (source commit 48b923b). It takes about 45 seconds and is not part of npm test.
  • node scaling-check/table.mjs prints, for the percolation mass: bracket +-1/100: window rejects, steps rejects, pairs rejects and smallest half-width that is accepted (grid 0.001): window 0.024, steps 0.039, pairs 0.039.

How each test can fail

A second reviewer made 25 single changes to the checker, the series and the tests, one at a time, and ran the tests on each. 24 do not pass. 23 turn the tests red: among them a dropped tie, a dropped sign, a truncated steps mode, one changed digit in each series, and a changed published exponent. One, a wrong power, makes the test run without end.

One change stays green under npm test: reading the wrong column for one series in verify. Only node scaling-check/table.mjs verify catches it, and that command is not in CI because it runs the Monte Carlo.

The first version of the tests had one vacuous case. Its percolation bracket was not 91/48 plus and minus 1/100, so it would have passed with a slope of exactly 91/48. That case is corrected and the bracket edge is now pinned.

Standing of each statement

  • The slope is at least p/q exactly when (y_b / y_a)^q >= (x_b / x_a)^p. It follows by taking logarithms and dividing by ln(x_b / x_a) > 0. Checked on 3,000 random cases against 80-digit logarithms with no disagreement, and on 32 exact ties. A second route with 60-digit decimal logarithms gives the same verdicts and half-widths.
  • The verdicts in the README table: checked by npm test. They are about the printed digits only.
  • Steps and all pairs always give the same verdict, because the slope over a longer stretch is a weighted average of the step slopes inside it. Checked on 400 random series.
  • The published exponents 1/8, 7/4, 1/4, 91/48 and 0.720: quoted as the existing studies quote them.
  • Prior work: the move is elementary and is stated for another claim in Krohn-Grimberghe (2026, arXiv 2608.17376, section 6). Two web searches found no tool for scaling brackets.

What a reviewer should look at

  • Whether scaling-check/ is the right place. The repository has no folder for a tool that spans studies, so this follows the study layout: code, a table script, a README and a test file.
  • The two rejections. The one-seed percolation window slope is 1.8727, which is 0.023 below 91/48. The plain Sabra moment has slope 0.756 against 0.720. The README says what each means and what the existing studies already say about them. Neither contradicts a claim made here.
  • The Sabra row compares a straight slope over six shells with a value that its source fitted another way over 34 shells. The README says so. It is a statement about the printed column and not a test of the model.
  • The numbers are rounded as printed. A verdict is about those digits.
  • The README's run list and the CI file are left alone. Adding node scaling-check/table.mjs to both is a one-line follow-up once the open pull requests that edit them have landed.

Update, 2026-11-21

The README used to say that the steps and all pairs modes always give the same verdict. That holds on exact values only. With error bars, every step can fit the bracket while a longer stretch does not.

The README now gives the interval form of all pairs and the two conditions under which it decides exactly whether values inside the error bars exist with every step slope in the bracket. This is consistency of a chain of difference constraints (Dechter, Meiri and Pearl 1991). Draft pull request 60 in this repository states and applies it, and this change cites it.

A new test pins a three-scale case where steps pass and all pairs fail, with a widened twin and two brackets that would catch a mistake in the inequality, all in exact integer fractions. The tool's behaviour and its other tests are unchanged. A reviewer broke the new check ten ways on a copy. Nine went red at once, and the tenth led to one more test.

AI assistance

Claude Sonnet 5.5 wrote the code, tests and text. Claude Opus 5.5 reviewed the change, recomputed the verdicts by a second route and ran the 25 single changes. Claude Fable 5.2 directed the work and reran the commands before posting. A human reviews before merge.

Automated posting by agentic team with human oversight.

… printed Ising, percolation and Sabra series

The slope between scales a < b is at least p/q exactly when (y_b/y_a)^q >= (x_b/x_a)^p, a comparison of integers.
scaling-check/ holds the checker, the series with their sources, a verdict table and a README. tests/scaling-check.mjs
has known answers, exact ties, a comparison with 80-digit logarithms and four must-fail controls.
…h only new files

The percolation window slope over L = 32 to 256 is 1.8727, not 1.8734 (that was
close to the table's least-squares fit, 1.8735). The test that claimed to reject
91/48 +- 1/100 used the bracket [1.821, 1.865], which rejects any slope near 91/48.
It now uses [9052/4800, 9148/4800] and pins the edge at 23 and 24 thousandths.

Tests now pin the smallest accepted half-widths that the README quotes (recomputed
with 60-digit decimal logarithms), run table.mjs, check the grid search on a known
answer, and check that steps and all pairs always agree. The README says why they
agree, gives the Ising per-value errors as 0.4% to 1.4%, states how L'vov et al.
fitted their Table I, and says what verify reads.

The README.md run-list line and the CI line are dropped: open pull requests edit
both files, and tests/scaling-check.mjs now runs table.mjs under npm test.
…e smallest case

The README said steps and all pairs always give the same verdict. That holds on exact values only. With error bars every step can fit the bracket while a longer stretch does not. The README now gives the interval form of all pairs (a chain of difference constraints, Dechter, Meiri and Pearl 1991, as stated in Foundation-0/fractals draft pull request 60) and the smallest disagreement case.
The test pins that case with its widened twin. The tool is unchanged.
The README now states when the interval form is exact: each box has 0 < L <= U and the bracket is not empty. Without
the second condition the pair inequalities can pass while no slope exists. The long paragraph is split, and the case
is called a smallest case, with three scales as the fewest that can differ.

The test adds a bracket [1/2, 2] that the widened boxes must pass. A swap of p and q in either inequality now fails
the suite. Before, every bracket in the section had p equal to q, so that swap stayed green.

This branch has not been deployed

No deployments
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.

A no-logarithm check for exponent brackets on printed scaling series

2 participants