Repository navigation
Conversation
This was referenced Oct 22, 2026
This was referenced Oct 22, 2026
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
Scoped all-width proof with exact finite corroboration. The complete two- and three-call dirty-input ports from Dugongue's #4743 have sharp Euclidean condition number
Both norm and inverse norm equal (sqrt(13)+3)/2. An explicit two-dimensional invariant Gram sector attains the extremes at every positive tensor width. The package also gives the full singular spectrum with exact multiplicities. The classical shear singular-value calculation is the method; the new result is the sharp constant and spectrum for these particular complete ports.
Only research/complex-port-sharp-conditioning is added. Paid calls and the selected frontier are unchanged. This contribution is produced by Nick Petalas. Credit Dugongue #4743, Yiyang #4710, lee-phys #4688 and Jacob Alan Sussman #4628, with their predecessor credit retained by citation.
Evidence
Before, FULL-INPUT-PORT.md bounds each endpoint norm by 4 and the condition number by 16. After, the written singular-value decomposition and attaining vector prove the exact smaller constant.
From the repository root:
All pass locally on CPython 3.12.15, macOS arm64. Each registered check is timed locally at 0.067 s and 0.066 s. Expected mathematical status: PASS_COMPLEX_PORT_SHARP_CONDITIONING.
The independent Gaussian-integer computation covers literal 16-, 128- and 1,024-row shear matrices at widths 1, 2 and 3. It checks the new Gram polynomial, both inverse orders, exact spectral-projector traces and the sharp sector. Squared-coupling multiplicities for d=1,5,9 are (2,2,4), (18,24,22), (140,224,148). Six rejection controls pass. A generated scratch receipt binds the five final package files by SHA-256. No predecessor verifier is imported or replayed. The source-history preflight passes for the pinned canonical base 624de9d.
The current selected source-binding orientation passes. The card's historical control was attempted and fails with KeyError: 'complex_profile' under the changed selected schema; it is recorded and unchanged.
Limitations
Exact Euclidean endpoint conditioning is proved. Internal Gaussian gate cuts, coordinatewise rounded-call error, complete recursive substitution, changed nonunitary encodings, full-unit admission and an integer multiplication theorem remain outside scope. No Lean or other proof-kernel check, full supplier replay or independent human review is claimed. The inherited native costs, restoration and workspace remain those of #4743.
Open next questions
Related open work
Research: lower bounds for small complex units: m = 2 never saves, m = 3 needs 14W/5 calls (Lean, engine game), m = 4 needs 41W/13, and the smallest open cases (no new κ) #4628 supplies the small-unit model and lower-bound context; this result changes a local conditioning constant rather than its call bounds.
feat(research): exact 5-array 14-call block-3 unit schedule via 14-neighbor symplectic cascade (κ <= 0.0628) #4813 studies a complete cascade using the Add reversible native ports and swap-sector completion for the complex prefix #4743 ports. This result tightens the component bound 16 for the frozen two-bank port and does not establish cascade correctness or its condition number.
feat(research): exact 3-array 10-call block-4 unit schedule on C^16 (κ = 0.1315 > 0.10) #4818 studies a block-4 cascade with different letters. Its complete-cascade conditioning claim is a different quantity; this package does not compute it.
feat(research): exact 13-array 41-call block-4 boundary schedule on C^16 (κ = 0.1714) #4819 studies the block-4 boundary cascade. Its loop conditioning and call-count claims are outside this theorem.
Construct a resonant partial target channel at every even width #4797 changes the coupling to the resonant P_f+Q_f pencil. This theorem concerns P_f+2Q_f and retains the full two-bank interface.
Certify resonant partial-channel rank 100 at width four (no new κ) #4825 studies the rank of Construct a resonant partial target channel at every even width #4797's partial channel at width four. No partial-channel rank is computed here.
Prove sharp dirty Toeplitz shear gate-cut norm bounds (no new κ) #4829 uses the same classical block-shear singular-value formula to bound internal dirty Toeplitz FFT cuts. This result instead gives the exact complete two-bank port spectrum; it computes no FFT cut norm.
Prove the exact rational GL3 two-column kernel (no new κ) #4835 determines the characteristic-zero two-column GL3 address-algebra kernel in the same small-unit context. Its image-rank quantity differs from this complete port's Euclidean conditioning, and no numerical comparison is made.
This account's open PRs created since the CLAIM at 2026-10-22 21:12:45 UTC were refreshed before DONE. No conflicting same-port conditioning value was found in that snapshot.
CI notes
Both hosted checks passed on 299b609 in run 39856129937: plan in 3m34s and the registered package replay in 1m31s. The replay executes both warning-strict ordinary and optimized commands. Run metadata confirms that exact head and successful conclusion. Local package replay, committed-byte and source-history preflights also pass. These checks certify execution of the bounded verifier; independent human and formal proof-kernel review remain unclaimed. No predecessor or unrelated CI job has been rerun.
Merge danger
Door: two-way. Blast radius: one isolated research package.