{"id":"b43dc15e-b116-45be-94ff-383f615be70f","arxiv_id":"2607.09339","paper_version":2,"verdict":"ACCEPT","confidence":"HIGH","novelty_score":6.0,"correctness_risk":"low","formal_verification":"none","parameter_count":0,"one_line_summary":"PCPOP.jl implements partially commutative polynomial optimization with algebraic, symmetry and Jordan reductions, delivering measurable speed and size gains on quantum-information SDP hierarchies.","lead":"PCPOP.jl is a Julia package that builds and solves semidefinite relaxations of polynomial optimization problems with non-commuting or partially commuting variables. It targets quantum-information tasks such as Bell inequalities and entropy bounds by exploiting commutation structure for smaller, faster programs.","discovery_kind":"new_method","skeptic_critique":{"model":"grok-4.5","headline":"No significant objection identified","rationale":"The reader correctly isolates the only non-trivial assumption (soundness of the clique normal forms) and correctly judges that it does not undermine an ACCEPT verdict for a software paper whose claims are checkable by installation and re-execution. The applications recover textbook optima, the algebraic micro-benchmarks of §6.1 show clear asymptotic gains over OSCAR and QuantumNPA, and the larger SDP tables document both smaller variable counts and faster solve times precisely on the partial-commutation-heavy instances for which the package was designed. No stronger load-bearing flaw (e.g., incorrect dual formulation, non-terminating reduction, or unfair baseline) appears in the manuscript. Hence the verdict and confidence remain unchanged.","tokens_in":34531,"tokens_out":484,"duration_ms":12744,"concrete_test":"Install PCPOP from the public registry, re-run the n-cycle contextuality instance of Table 4 (n=5, d=4) and the CHSH first-level relaxation of §4.3/§5.1 under both primal=true and primal=false; confirm that the optimal values match the published figures to solver tolerance and that the SDP sizes remain strictly smaller than the corresponding QuantumNPA/Moment instances.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim is a software-performance claim: the clique-representation arithmetic (with built-in projector/unitary/unipotent reductions and the linear-time cyclic test of [34]) yields valid, smaller SDPs that are faster to build and solve than generic Gröbner or substitution approaches on QI problems. Correctness of the normal forms is deferred to the companion theory paper [40] and is not re-proved here; if those forms ever failed to identify an equivalence or introduced a spurious one, the resulting moment matrices would not be valid relaxations. In the present manuscript that risk is mitigated by extensive recovery of known tight values (CHSH = 2√2, KCBS, Mermin bilocal, etc.) and by the fact that the package ships with exact-arithmetic support and reproducible tutorials. No internal inconsistency, circular construction, or benchmark artifact that would overturn the reported speed-ups was found.","agreement_with_reader":"agree"},"referee_report":{"model":"grok-4.5","summary":"The manuscript presents PCPOP.jl, a Julia package for building and solving moment/SOS semidefinite relaxations of non-commutative, tracial, trace, and state polynomial optimization problems. Its distinguishing contribution is a specialized partially commutative monoid implementation (clique/graph-product representations with built-in projector, unitary, unipotent, and orthogonality reductions, plus a linear-time cyclic-equivalence test) aimed at quantum-information problems. The paper supplies mathematical background (§1), feature and implementation descriptions (§2–3), a tutorial with runnable examples (§4), a broad applications chapter (Bell scenarios, contextuality, conditional entropies, networks, uncertainty relations, almost qudits, information capacity; §5), and benchmarks against Ncpol2sdpa, QuantumNPA, Moment, and OSCAR (§6).","tokens_in":34740,"tokens_out":1291,"duration_ms":17103,"significance":"If the reported normal forms are correct and the package is maintained, this is a useful, practical contribution for the quantum-information community: it unifies several NPA-style hierarchies in one Julia interface, adds exact-arithmetic and symmetry/Jordan reductions, and—most importantly—exploits partial commutativity in a way that yields smaller, faster SDPs on standard QI benchmarks (notably n-cycle contextuality and related scenarios in Tables 3–4). Strengths include reproducible tutorial code, recovery of known tight values (CHSH, KCBS, bilocal Mermin, etc.), side-by-side SDP-size and timing tables against independent packages, and an open feature comparison (§2.7). Correctness of the core clique arithmetic is appropriately deferred to the companion theory paper [40] and the cyclic algorithm of [34]; the present manuscript’s role is software design, interface, and empirical performance.","major_comments":[{"comment":"The central performance claim rests on the clique/graph-product normal forms (and built-in projector/unitary/unipotent reductions) producing valid, complete equivalence classes for the moment matrices (§2.1, §3.2). Correctness is deferred to [40] and [34]. The manuscript should state this dependency more explicitly near the main claim (abstract/§2.1) and list, in one place, the known tight values recovered in §5 as empirical validation that the resulting SDPs remain valid relaxations. Without that short, load-bearing clarification, a reader cannot assess the risk that a missed or spurious identification would invalidate a relaxation.","section":null},{"comment":"Reproducibility of the software claim: the installation snippet (§4.1) uses Pkg.add(\"PCPOP\"), but the manuscript does not give a version pin, repository URL/DOI, or commit corresponding to the §6 timings (20-core i7-12700, 64 GB). For a package paper whose main result is comparative performance (Tables 1–4), a fixed release identifier and a short note on solver versions (Mosek) and floating-point vs exact modes used in the tables are needed so that the reported speed-ups can be re-run.","section":null}],"minor_comments":[{"comment":"Several tutorial/application snippets leave variables undefined or unused (e.g. §5.4 prints val without assigning objective_value(model); §5.3 mixes npa_dual/model_new_obj with the pcpop API of §4.2). Align examples with the documented public API or mark internal helpers.","section":null},{"comment":"§4.3 CHSH code uses a[2,0] b[2,0] with Unipotent and objective a[1]*b[1]+… while the displayed problem (25) indexes a0,a1,b0,b1; the indexing is consistent in 1-based Julia but the prose/equation labels should match the code arrays to avoid confusion.","section":null},{"comment":"Table 1 caption says SDP size, #cons, and #vars are the same for all packages, yet the narrative in §6.2 notes implementation differences in how equivalence constraints are counted (PCPOP vs others). Clarify when counts are forced equal vs when PCPOP’s single-PSD formulation reduces #cons (as in Table 3).","section":null},{"comment":"§2.7 compares features with NCTSSOS and SumOfSquares but §6 benchmarks only Ncpol2sdpa, QuantumNPA, Moment, and OSCAR. A one-sentence note that NCTSSOS sparsity was out of scope for this release would prevent a missing-baseline objection.","section":null},{"comment":"Typos and polish: “Gr¨ obner” spacing throughout; “semidefinitze” (§4.10); “optimzation” (§5.1); “Paw/suppress lowski” and similar bibliography encoding artifacts; “Beno ˆ ıt” in acknowledgements. Clean UTF-8 and LaTeX accents.","section":null},{"comment":"Figure 1–3 axes use Time (ns) with log scale; state the number of random trials and whether times include monoid build or only multiply+reduce, so the OSCAR vs PCPOP comparison is unambiguous.","section":null},{"comment":"§1.6 lists three SDP encodings; default choices (primal=true, canonical matrix variables) are stated later in §4.2. Cross-reference defaults once in §1.6 for readers who skip the tutorial.","section":null}],"recommendation":"minor_revision","confidential_remarks":"This is primarily a software/manual paper whose novel algorithmic content lives in the companion theory work [40]. For a quant-ph venue that accepts computational tools, that is acceptable if the package is public and the benchmarks hold; if the journal expects self-contained theory, the fit is weaker and the editor may prefer a software-focused track or a short companion note. I did not find internal inconsistency or benchmark artifacts that would overturn the reported advantages on the QI instances shown."},"author_rebuttal":null,"desk_editor":{"model":"grok-4.5","letter":"This is a software contribution that does what it claims. The new piece is the clique-representation / graph-product monoid machinery (with the edge_l / edge_r bookkeeping and the built-in projector/unitary/unipotent reductions) that turns partial commutativity into a first-class, efficient representation rather than a post-hoc substitution rule. That is the right engineering choice for the large class of Bell, contextuality, network and prepare-and-measure problems that dominate quantum-information applications of non-commutative polynomial optimization.\n\nWhat the paper does well is concrete and checkable. Sections 3–4 give the data structures and the multiplication/division/cyclic algorithms in enough detail that you can see how the reductions are performed; the tutorial and application sections recover the known tight values (CHSH = 2√2, KCBS, Mermin bilocal, etc.); and the benchmarks in §6 show real reductions in SDP size, variable count and solve time against Ncpol2sdpa, QuantumNPA and Moment on the same problems. Exact arithmetic, Wedderburn symmetrization and Jordan reductions are also present and usable. The comparison table with existing packages is fair.\n\nThe soft spot is the one the stress-test already flags: correctness of the normal forms is deferred to the companion theory paper [40] and the linear-time cyclic test of [34]. That is normal for a software paper, and the extensive recovery of known optima plus the open code make the risk low, but a referee will still want a short, self-contained argument or a pointer to the precise theorem that guarantees the clique projections identify exactly the right equivalence classes. No other load-bearing gaps appear; the free-parameter and circularity burdens are essentially zero.\n\nThis is for people who already solve NPA-style hierarchies or who need state/trace polynomials on partially commuting operators. It is not a theory paper and does not pretend to be. I would bring it to reading group if we have anyone actively coding these relaxations, I would cite the package when I next need a partial-commutativity-aware solver, and I would send it to peer review without hesitation. Accept after the usual software-paper checks (install, re-run a couple of examples, confirm the benchmark numbers).","headline":"Solid, usable software paper: specialized partial-commutativity data structures that measurably shrink and speed up NPA-style SDPs for the QI problems that actually use them.","tokens_in":35311,"tokens_out":563,"would_cite":true,"duration_ms":6488,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["90C22","13P10","81P45"],"pacs":["03.65.Ud","03.67.-a"],"model":"grok-4.5","headline":"A Julia package turns partial commutations into faster, smaller SDP relaxations for quantum information problems.","keywords":["partially commutative polynomials","noncommutative polynomial optimization","semidefinite relaxations","quantum information","Bell inequalities","state polynomials","trace polynomials","Julia package"],"falsifier":"On any of the benchmark problems (n-cycle CHSH, bilocal Mermin, conditional entropy bounds) construct the same level of the hierarchy with a complete Gröbner basis or an independent package and check whether the optimal values differ beyond solver tolerance, or whether an intermediate product of two clique words fails to equal the product computed by free monoid reduction.","tokens_in":35448,"feed_emoji":"⚛️","tokens_out":817,"duration_ms":8845,"temperature":0.7,"pith_summary":"PCPOP is a Julia package that builds and solves the standard hierarchies of semidefinite relaxations for polynomial optimization, covering ordinary non-commutative, tracial, trace and state polynomials. Its distinguishing mechanism is an arithmetic that works directly with partially commuting variables: words are stored by their projections onto maximal cliques of the dependence graph, so multiplication, involution, equality and cyclic equivalence become cheap operations that automatically absorb projector, unitary, unipotent and orthogonality relations common in quantum information. The same representation yields smaller moment matrices than generic Gröbner or substitution methods. The package also automates Wedderburn symmetry reduction, Jordan-algebra dimension reduction and exact-arithmetic Gröbner bases, and is illustrated on Bell inequalities, contextuality, networks, uncertainty relations and device-independent cryptography. Benchmarks against existing tools show both faster set-up and smaller SDPs on the problems that exploit partial commutativity.","feed_headline":"Partial commutations cut the cost of quantum SDP hierarchies","feed_subtitle":"Clique arithmetic yields smaller, faster moment relaxations for Bell, network and crypto problems","key_machinery":"Clique representation of partially commutative words: each word is stored as its projections onto the maximal cliques of the dependence graph; multiplication, involution, division and cyclic equivalence then reduce to ordinary non-commutative operations on those projections, automatically incorporating the algebraic relations that appear in Bell and contextuality scenarios.","core_discovery":"The central claim is that a clique-representation arithmetic for partially commutative monoids, together with built-in reductions for projectors, unitaries, unipotents and cyclic equivalence, produces valid and substantially smaller semidefinite relaxations for the non-commutative, tracial, trace and state polynomial optimization problems that arise in quantum information.","pith_inferences":["The same clique arithmetic could be grafted onto existing sparsity or chordal-completion techniques to obtain hybrid reductions that neither package currently offers.","Because cyclic equivalence is decided in linear time, tracial and free-probability hierarchies become practical for alphabets that previously forced exponential Gröbner bases.","If the companion theory paper’s correctness proof holds, any future quantum-information SDP that can be written with partial commutations and projector relations can safely drop its hand-crafted substitution tables."],"forward_implications":["Standard Bell, contextuality and network hierarchies can be built at higher levels or with more parties before the SDP becomes intractable.","Device-independent entropy and key-rate bounds that previously required custom substitution rules become one-line calls that exploit the native partial-commutativity arithmetic.","Exact-arithmetic pipelines (Gröbner + Wedderburn + rational SDP rounding) become routine for the same quantum-information problems.","Graph-product monoids give a compact algebraic language for multi-partite scenarios with overlapping measurement supports."],"fun_headline_variants":["Clique arithmetic shrinks quantum SDP hierarchies for Bell problems","Partial commutations yield smaller moment relaxations in quantum info","PCPOP cuts cost of noncommutative poly optimization via monoid cliques","Built-in reductions produce faster SDP relaxations for quantum networks","Partially commutative monoids enable leaner tracial polynomial SDPs"],"cache_read_input_tokens":32896,"weakest_assumption_plain":"That the clique projections plus the built-in projector/unitary reductions correctly identify every pair of words that are equivalent under the given partial commutations, so the resulting moment matrix is still a valid relaxation of the original problem.","fun_headline_variants_meta":{"raw":{"variants":["Clique arithmetic shrinks quantum SDP hierarchies for Bell problems","Partial commutations yield smaller moment relaxations in quantum info","PCPOP cuts cost of noncommutative poly optimization via monoid cliques","Built-in reductions produce faster SDP relaxations for quantum networks","Partially commutative monoids enable leaner tracial polynomial SDPs"]},"model":"grok-4.5","effort":"low","cost_usd":0.003958,"raw_usage":{"total_tokens":1133,"prompt_tokens":614,"num_sources_used":0,"completion_tokens":90,"cost_in_usd_ticks":39580000,"prompt_tokens_details":{"text_tokens":614,"audio_tokens":0,"image_tokens":0,"cached_tokens":256},"completion_tokens_details":{"audio_tokens":0,"reasoning_tokens":429,"accepted_prediction_tokens":0,"rejected_prediction_tokens":0}},"tokens_in":614,"tokens_out":90,"duration_ms":3447,"temperature":1.0,"reasoning_tokens":429,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-07-14T15:13:37.662531+00:00","model_set":{"reader":"grok-4.5"},"falsifier":"On any of the benchmark problems (n-cycle CHSH, bilocal Mermin, conditional entropy bounds) construct the same level of the hierarchy with a complete Gröbner basis or an independent package and check whether the optimal values differ beyond solver tolerance, or whether an intermediate product of two clique words fails to equal the product computed by free monoid reduction.","supporting_citations":[],"review_version":2}