Loss decoding based on Pauli envelope by yuewuo · Pull Request #166 · microsoft/qdk-ec

Yue Wu (yuewuo) · 2026-08-19T01:49:41Z

Adds end-to-end Pauli-envelope loss decoding support:

Compiles configurable gate-loss policies into causal loss graphs while preserving herald identity across gadgets and decode windows.
Adds language support for the Pauli-envelope data that can is output by the annotator and can be inspected or edited by user
Adds coordinator strategies to ignore loss, convert it into shot-scoped edge reweights, or pass structured LossInfo to loss-aware decoders.
Supports decoder capability negotiation, persistent graph reuse, and random imputation of lost measurement outcomes.
Includes a herald-aware MLE reference decoder, updated protobuf/Python/CLI APIs, comprehensive tests, and tutorial documentation.

…se-decorator

Merges 8 commits from main including PR #84 (@REPROPAGATE). Conflict resolution: - compose_builder.py: kept HEAD (local @REPROPAGATE implementation is a superset of main's; includes _translate_compose_conditionals). - jit_annotate.py: kept HEAD (CONDITIONAL R<j> emission block). - jit_noise_builder.py: merged error message wording (main's 'or' + HEAD's extra_clause). - teleportation.deq fixture: kept HEAD (additional Teleporatation2 test compose). - test_compose_repropagate.py: kept HEAD (added TestTranslateComposeConditionals class). - jit_library_builder_test.py: kept both (HEAD's 8 compose_conditional tests + main's 3 build_jit_program tests are non-overlapping). Known integration issues (to fix in follow-up commits): - 9 tests fail due to divergent @REPROPAGATE implementations between HEAD's f526417 and main's PR #84. The basis-freedom check in jit_noise_builder.py:1788 doesn't see row_extra_descs for CONDITIONAL contributions in the merged code. - 2 runtime test files fail to collect (deq_runtime needs maturin rebuild for main's new APIs).

Co-authored-by: Yue Wu <yuewu4@microsoft.com>

Co-authored-by: copilot-swe-agent[bot] <198982749+Copilot@users.noreply.github.com>

…t, fix swapped Clifford gates (#100) Co-authored-by: jbellorivas <jbellorivas@microsoft.com> Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

Co-authored-by: Juan M. Bello-Rivas <jbellorivas@microsoft.com> Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

…er (#172) Signed-off-by: dependabot[bot] <support@github.com> Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com>

Signed-off-by: dependabot[bot] <support@github.com> Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com>

A formal model of quantum error correction protocols. Authored-by: Adam Paetznick <adpaetzn@microsoft.com>

…lop-loss # Conflicts: # .ado/stages/build.yaml # .github/workflows/build.yaml

Use union rather than parity across parent-child loss sites. Keep parity unchanged for independent toggles. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com> (cherry picked from commit dc5f0a4eadab29b26dafd9fd82c7a34fef2f5bef)

Update the overlap expectation and add the certain-parent/child regression. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

The loss-aware ILP required a hyperedge to be selected whenever it arose from an ordinary error. Pauli-envelope semantics instead combine ordinary-error and loss contributions with exclusive or, so matching contributions cancel. Allow the ordinary and loss contributions to cancel when the loss is selected. Add regressions showing that this restores feasibility for the probability-one example and makes the subunit example choose no net edges instead of the three-edge cycle. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

Document the feasible endpoint assignment and the finite-probability cost comparison directly beside their regressions so reviewers do not need the external audit proofs. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

Describe each regression using only its graph, probabilities, detector parity, and expected decoder output. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

Juan M. Bello-Rivas (jmbr) · 2026-09-18T21:40:11Z

        for (size_t ei : best) {
            for (size_t orig = 0; orig < original_error_map.size(); ++orig) {
                if (original_error_map[orig] == ei) { result.push_back(orig); break; }
            }
        }


Tesseract merges hyperedges that flip the same detectors: a hyperedge coming from the Pauli envelope with probability zero might be merged with another identical hyperedge having nonzero probability. The Tesseract interface may return the hyperedge with probability zero because the mapping from a merged hyperedge back to the original hyperedges ignores their probabilities.

Fixed by ff39a8b

Tesseract previously mapped every selected merged error to its first original member, even when that edge's current probability was zero. Preserve original costs across construction and reweighting, then track the lowest-cost member of each merged group. The regression covers loaded probabilities, shot-scoped reweighting, and restoration of the loaded probabilities. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

Juan M. Bello-Rivas (jmbr) · 2026-09-18T22:02:35Z

+            self.use_loaded_reweights,
+        )
+        .await
+        .unwrap();


I think the coordinator at this stage does not have to panic because decode_projected will return a Result that could be handled. I think the same happens earlier in this coordinator and in the monolithic one.

In general, I remain uneasy about the usage of unwrap (see last paragraph of the section Shortcuts for Panic on Error).

Yue Wu and others added 30 commits June 18, 2026 10:36

add better error message

d8ae687

add support for REPROPAGATE and --keep-noise

426ebf0

Merge remote-tracking branch 'origin/main' into feature-issue-79-como…

89aaab0

…se-decorator

add tutorial check

3d09909

add tutorial chapter

afa25af

add support for conditional

20bfddf

fill test gap

ba49686

update teleportation part

f526417

update test deq files

457eb9a

add jit test

ee1268f

update old tutorial chapters to be consistent

552f699

split conditional correction chapter into two

2272f80

add figures

b6d13d1

update language

863aae0

fix language

b1e22b7

update test files

12adcf7

update

6dedd72

update language

c9cdc59

update tutorial chapters

3857d09

check in lattice surgery chapter

d915a55

update test reference files

2a6b41a

update tutorial chapters

e186a17

add readout propagation chapter

79d03be

add generator script

3a13502

add changes

545f604

update proto binding

6fa15c8

fix pipeline failure (#96)

748e8ef

Co-authored-by: Yue Wu <yuewu4@microsoft.com>

Cap grpcio-tools below 1.82 to fix protobuf gencode/runtime drift (#103)

5a749ec

Co-authored-by: copilot-swe-agent[bot] <198982749+Copilot@users.noreply.github.com>

pauliverse/paulimer: expose FramePropagator to Python, add reset_qubi…

52513c6

…t, fix swapped Clifford gates (#100) Co-authored-by: jbellorivas <jbellorivas@microsoft.com> Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

Yue Wu and others added 14 commits September 16, 2026 14:18

add test case

4a729c3

format

bdc4d23

avoid out of sync constants

76ba601

move some debug-only asserts to always

af84c07

Pin GitHub Actions to full-length commit SHAs (#171)

151e536

Update PyO3 dependencies to 0.29 (#173)

ebd9171

Co-authored-by: Juan M. Bello-Rivas <jbellorivas@microsoft.com> Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

Bump postcss-selector-parser from 7.1.1 to 7.1.5 in /deq/deq_visualiz…

93307f0

…er (#172) Signed-off-by: dependabot[bot] <support@github.com> Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com>

Bump @humanfs/node from 0.16.7 to 0.16.8 in /deq/deq_visualizer (#174)

c0789e4

Signed-off-by: dependabot[bot] <support@github.com> Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com>

Hello, qodec. (#178)

d8588df

A formal model of quantum error correction protocols. Authored-by: Adam Paetznick <adpaetzn@microsoft.com>

Merge remote-tracking branch 'origin/main' into feature-90-pauli-enve…

9bc7c16

…lop-loss # Conflicts: # .ado/stages/build.yaml # .github/workflows/build.yaml

remove HTML deq code highlights

6797e22

fix double counting of source and continuation errors

2ba2ade

fmt

15ef82a

fix bug of --detectors panic in transpile CLI

59bbe63

Yue Wu (yuewuo) requested a review from Juan M. Bello-Rivas (jmbr) September 17, 2026 04:30

Yue Wu added 4 commits September 17, 2026 09:25

bump rc version

aac0cbf

fix paulimer build error

6af1efb

upgrade ziglang version

022974a

pin crc32fast version to avoid panic in MS windows Rust compiler

b4b5ca6

Juan M. Bello-Rivas (jmbr) requested changes Sep 18, 2026

View reviewed changes

Comment thread deq/deq_runtime/src/coordinator/loss_handler.rs

Juan M. Bello-Rivas and others added 3 commits September 17, 2026 19:00

test(loss): cover causal overlap activation

96c5890

Update the overlap expectation and add the certain-parent/child regression. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

update version to 0.5.0

8968d02

Juan M. Bello-Rivas (jmbr) reviewed Sep 18, 2026

View reviewed changes

Comment thread deq/deq_runtime/src/decoder/mle_loss_decoder.py

Juan M. Bello-Rivas and others added 3 commits September 18, 2026 13:25

test(loss): make cancellation rationale self-contained

21ce4da

Describe each regression using only its graph, probabilities, detector parity, and expected decoder output. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

Juan M. Bello-Rivas (jmbr) reviewed Sep 18, 2026

View reviewed changes

Navigation Menu

Loss decoding based on Pauli envelope - #166

Loss decoding based on Pauli envelope#166
Yue Wu (yuewuo) wants to merge 197 commits into
mainfrom
feature-90-pauli-envelop-loss

Yue Wu (yuewuo) commented Aug 19, 2026 •

edited

Loading

Uh oh!

Uh oh!

Uh oh!

Juan M. Bello-Rivas (jmbr) Sep 18, 2026

Uh oh!

Juan M. Bello-Rivas (jmbr) Sep 18, 2026

Uh oh!

Juan M. Bello-Rivas (jmbr) Sep 18, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

5 participants

Conversation

Yue Wu (yuewuo) commented Aug 19, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Juan M. Bello-Rivas (jmbr) Sep 18, 2026

Choose a reason for hiding this comment

Uh oh!

Juan M. Bello-Rivas (jmbr) Sep 18, 2026

Choose a reason for hiding this comment

Uh oh!

Juan M. Bello-Rivas (jmbr) Sep 18, 2026

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

5 participants

Yue Wu (yuewuo) commented Aug 19, 2026 •

edited

Loading