Regrouping · Iteration 03 · Current discussion

A fresh dimension for each assignment

Clarification recorded · local research checkpoint 7df84ea.

Christopher questioned whether talk of “height” had narrowed his idea to 3D. His proposal permits another dimension at every square-placement step, with an exact inverse and no fixed global dimension cap.

Illustrations added to this record . The original five captures and the later KR1 clarification diagrams are included below; click any image to enlarge it.

Objects and cancellation clarified · 2026-10-25T19:20:07Z

What genuine KR1 diagrams look like

Christopher asked to see actual residue and gap partitions, and whether a rearrangement could stay in 2D. These are unfilled Young/Ferrers diagrams: row lengths are parts and each square contributes one unit. Staying in 2D is allowed; validity, weight, inverse and complete coverage are the obligations.

Two genuine weight-30 KR1 partitions: residue parts 10,8,6,3,3 and ascending gap parts 2,4,5,8,11. The close pair 4,5 is highlighted.
Two valid equal-weight examples, not an asserted pair under a general bijection. Residues are 1,8,6,3,3 modulo 9. The distance-two differences in the gap partition are 3,4,6; its close pair 4+5=9 is divisible by 3. KR1 conditions in Mizuno’s introduction.
Explore the weight-30 and weight-21 examples

The gap rules concern ascending positive parts: parts two positions apart differ by at least 3; adjacent parts differing by at most 1 must sum to a multiple of 3. No cell labels or graph-inversion statistic are part of this original KR1 definition.

Cancellation is a separate signed-tableau problem

A determinant expansion introduces plus and minus occurrences. Each contributes a signed term ±qinv_G(T). An opposite-sign pair with the same graph-inversion count contributes zero. This removes algebraic contributions, not squares from a KR1 partition.

A valid negative tableau (1,3)/(2,4)/(5) and positive tableau (1,3)/(2)/(4)/(5), each with one graph inversion, contribute minus q and plus q, which cancel.
An actual pair for the signed e_(4,1) coefficient at n=5, with h=(1,3,3,4,5). The incomparability graph has only edge {2,3}. In both tableaux, 3 lies in an earlier row than 2, so inv_G=1. The signs come from their determinant sectors; every cell still has an ordinary positive label.
T1 and T2 have different placements of 4 and 5, but the coarse slide gives one common positive hook. Their displayed signed subtotal is plus one minus one minus one, leaving minus one.
The original T1/T2 collision in the same valid poset: all three displayed occurrences have inversion count zero. One positive occurrence cancels only one negative occurrence. The other needs a different allowed positive partner. The −1 is the remainder of this displayed group, not the value of the complete coefficient.
Switch between the valid cancellation and the collision

The primary KR1 task remains a reversible, weight-preserving translation between all original residue and gap partitions. Repairing this separate tableau slide alone would not supply that translation.

Illustrations from this regroup

2D and 3D: remember which numbered square moved

T2 has 5 in its corner and 4 at the bottom. Sorting loses that choice; a named z axis retains the corner square 5.
The 2D sorted output loses the starting arrangement. The restricted 3D corner lift retains square 5 on the named z axis, with square 4 still on y.
4D displayed as two slices: w=0 contains 1,2,3 and 5 on z; w=1 contains square 4.
4D is shown as separate w=0 and w=1 slices. This illustrative next move remembers 5 on z and the former bottom square 4 on w.

The 3D corner lift is a reviewed restricted construction. The 4D move is an illustrative extension, not a general KR1 target map.

Christopher Yamba's intuition

Let the arrangement retain each selection

“Each step is an assignment of a tableaux square and number and you can extend each step into a new dimension.”

Place the actual numbered squares in a discrete multidimensional configuration. Each step may introduce another named axis. Record the selected move, recover it from the endpoint, and compare working and failing families for a uniform pattern.

Different finite inputs may use different finite dimensions. We seek existence of one suitable procedure; neither a unique procedure nor an optimal dimension is required.

Scope correction: the latest KR1 attempt mainly explored column counts and carries. It did not implement this full per-square procedure. The reviewed 3D lift is a restricted case; the final-target height condition is not a cap on intermediate dimension.

A fresh axis for each assignment

Step zero of the fresh-axis placement procedure: T2 and its exact original x,y coordinates for numbered squares 1 to 5.
The displayed initial state is 2D. In the interactive version, each Next assignment moves one square along a fresh named axis; five assignments reach 7D. Undo reverses the last step. The original x,y coordinates remain part of the state.
Try T1 / T2 and the reversible placement steps

Switch T1 and T2 to follow the swapped 4 and 5. Use Next assignment and Undo assignment to inspect the coordinate record.

What this example establishes: a recoverable placement mechanism on the illustrated configurations. The 4D move and the lower stepper are separate illustrative conventions. The stepper allows sparse states, so it is not a proof of Ferrers admissibility or of all required statistics in general.

The bridge the algorithm still needs

The second correspondence need not be literal geometric flattening. It must preserve the required weight, distinguish inputs both within and across fibers, cover exactly the required target family, and let the final object determine every reverse choice. Simply forgetting the extra coordinates may restore the collision.

We must also decide which intermediate states are allowed. A genuine finite Ferrers down-set has cell-budget restrictions on active axes; a sparse working arrangement is a different intermediate object. The target translation must account for that choice.

A concrete KR1 test: memory and the bottom boundary

The accepted carries retain a recognizable exchange count and small columns that keep the terminal boundary valid. The failed exchange preserves weight but breaks that boundary. This is a test for any proposed decoder from the extra-dimensional arrangements.

Accepted weight 48 carry: columns 12,9,9,6,6,3,3 become 12,9,9,9,6,3, retaining a bottom drop of 2.
Before and after the accepted weight-48 carry: one 6 and one 3 become a 9. A 6 and a 3 remain, so the bottom drop stays 2.
Second captured view of the weight 48 carry, showing the encoded values and highlighted bottom drop.
Christopher’s additional captured view of the same carry. The numbers above the columns are decoded heights; the encoded column values below sum to 48. Equal total weight alone does not establish admissibility.
Compare the working carry with the failed exchange
Patterns and the next handoff

Record the puzzle, then state the rule

The next proposal should specify literal placements, intermediate admissibility, invariant, selected endpoint language and a target-only decoder before checking. Use the remaining h000 core case from iteration 02 as the first KR1 readout testbed.

Record each step with a full timestamp, label, old/new coordinates, introduced axes, selected permutation, invariant, inverse guard and first failure. The external log helps find patterns; the final inverse must not depend on a missing diary.

Current status: this is a proposed general approach with reviewed restricted ingredients and illustrative traces. The full KR1 bijection remains unresolved. This regroup has not launched another research run.