Repository navigation
Compile the width-13 Clifford decoder into 4,265 Bell reads and certify its dense input cost (no κ) - #7800
Open
apoursh wants to merge 1 commit into
Open
Compile the width-13 Clifford decoder into 4,265 Bell reads and certify its dense input cost (no κ)#7800apoursh wants to merge 1 commit into
apoursh wants to merge 1 commit into
Conversation
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Result
This takes the implementation-cost handoff from #7799. The width-13 Clifford decoder compiles into 4,265 distinct signed Bell-block reads per symplectic fibre, shared by its 256 Pauli coefficients. For the identity coefficient, on arbitrary dense computational-basis observations, the exact number of required scalar reads is 151 × 2⁴⁷ = 21,251,360,741,654,528. Compiler and examples · Group certificate · Raw acquisition and tape costs
Grouping the shear characters by their actual three-plane and quotienting repeated cosets reduces the source's double character sum. Direct expansion checks all 34,120 signed terms on 8,304 distinct symplectic elements. Only the target survives. The package emits the block addresses, exact weights and transport phases. An independent 64-operator mixture test recovers 512 coefficients for two fibres.
The identity traces share 22 shift geometries. Their frequencies use at most ten independent binary directions per geometry. Exhausting all 10,530 parity assignments proves that the combined raw coefficients have two nonzero magnitudes, 1/32 and 1/8. The executable supports random lookup and an address generator which enumerates each required input exactly once. The dense-input lower bound follows by perturbing one unread coordinate. Each normalized coefficient row has squared Euclidean norm 2⁴⁵. For the whole decoder we obtain an operator norm at most 4¹²√(1319/64), below 2²⁷.
These bounds apply to the emitted realization. Coalescing equivalent coset observations preserves recovery on the Clifford-mixture image and specifies an extension on arbitrary noise. Other decoder rows and compact or pretransformed inputs can have different access costs. The record-tape result counts input-head moves separately from arithmetic and workspace movement. No physical supplier or improvement to κ is claimed.
Verification
All eight registered runs passed on CPython 3.12.14. They include normal and optimized runs of the compiler, raw address checks, independent literal tensor checks and mixture recovery. Standard library only.
The raw verifier checks 21,060 cell-boundary lookups, the exact row norm, the extreme addresses and normalized diagonal Clifford column recovery. Separate literal tests check 262,144 shear phases, 24,576 transvection columns and six sparse raw inputs covering both signs, both magnitudes and cancellation. The enormous dense input is counted algebraically. It is never materialized.
Committed bytes, modes, source history and the official export passed. All eight checks also passed from a read-only Git-free export with an empty PATH. Local verification is independent of hosted workflow status.
Collaboration and credit
Base and prepublication refresh were canonical main
e54523517d0f03e58352e647c403a5af3e525a57. The investigation follows the scoped implementation handoff on the revision-8 audit track. Prepared by apoursh with substantial OpenAI Codex and separate agent assistance, under Apache-2.0. Credit the width-13 construction in #7799, Ali Romman's Bell-sector work, and the preceding character and Pauli work of Nick Petalas and alephnan. Collaboration claim · Sources