You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
docs: verify and upgrade all documentation for v0.6.2 (full Mac Intel+ARM support, hypergraph rejection details, sdist, accurate limits, install examples, roadmap reality)
Copy file name to clipboardExpand all lines: README.md
+1-1Lines changed: 1 addition & 1 deletion
Display the source diff
Display the rich diff
Original file line number
Diff line number
Diff line change
@@ -342,7 +342,7 @@ The REST API is for local experiments, partner review, or controlled internal de
342
342
| GPU performance | Correctness is artifact-backed for tested machines. Speedup is not universal. |
343
343
| OpenCL wheels | OpenCL support depends on build configuration and target environment. Confirm locally. |
344
344
| SparseBlossom | Near-optimal, not exact MWPM. Use `BlossomDecoder` for exact minimum-weight matching. |
345
-
| UnionFind | Fast approximate path; not a universal decoder for arbitrary graphs. |
345
+
| UnionFind / FastUnionFind | Fast approximate; only for codes where each qubit participates in ≤2 checks (graph-like matching codes). Explicitly rejects hypergraphs (v0.6.2+). Use Blossom/SparseBlossom/BPOSD for general cases. |
346
346
| REST/gRPC/MCP surfaces | Not hardened as public SaaS without a separate deployment/security review. |
Copy file name to clipboardExpand all lines: docs/CORRECTNESS_AUDIT.md
+2-2Lines changed: 2 additions & 2 deletions
Display the source diff
Display the rich diff
Original file line number
Diff line number
Diff line change
@@ -25,8 +25,8 @@ check matrix.
25
25
|---|---|---|---|
26
26
|`BlossomDecoder`| Syndrome-faithful and exact minimum-weight on audited small matching codes | Exhaustive brute-force oracle + PyMatching cross-checks | Exact MWPM on the tested audited graph families; do not claim universal hardware layout equivalence |
27
27
|`SparseBlossomDecoder`| Syndrome-faithful and near-optimal | Brute-force small-code audit + PyMatching-compatible matching-graph tests | Region-growing decoder, not exact MWPM by design |
28
-
|`UnionFindDecoder`| Syndrome-faithful on supported QEC matching graphs | Exhaustive small-code tests + family tests + property tests | Fast approximate decoder; not minimum-weight and not guaranteed on arbitrary adversarial hypergraphs|
29
-
|`FastUnionFindDecoder`| Same as Union-Find, lower-overhead path | Cross-decoder faithfulness tests | Fast approximate decoder, not exact MWPM|
28
+
|`UnionFindDecoder`| Syndrome-faithful on supported QEC matching graphs (each qubit in ≤2 checks) | Exhaustive small-code tests + family tests + property tests + hypergraph-rejection tests (v0.6.2) | Fast approximate; explicitly rejects hypergraphs (qubit degree >2) with clear error since v0.6.2. Not minimum-weight.|
29
+
|`FastUnionFindDecoder`| Same as Union-Find (≤2 checks per qubit), lower-overhead path | Cross-decoder faithfulness + rejection tests (v0.6.2) | Same rejection and limits as UnionFindDecoder.|
30
30
|`LookupTableDecoder`| Exact/faithful on stored table entries; faithful fallback on larger cases | Exhaustive table tests + d=5 fallback test | Table size and fallback behavior must be stated |
31
31
|`BPOSDDecoder` / `BpOsdDecoder`| Syndrome-faithful on LDPC/qLDPC-style CSS checks | BP-OSD reference-package comparison + CSS commutation tests | Quote LER from the harness, not from faithfulness alone |
32
32
|`BeliefMatching`| Lower observed LER on selected correlated workloads | Seeded Stim/PyMatching comparison tests and benchmark artifacts | Accuracy mode, not fast path; selected workload only |
0 commit comments