English HTML Slides
Prepared from arXiv v1 · May 29, 2026

Title

SAT, MaxSAT, and SMT for QLDPC Distance Computation

A presentation-style walkthrough of the paper’s motivation, formulations, benchmark design, and the main empirical takeaways for exact QLDPC distance computation.

Yu-Fang Chen · Seyed Mohammad Reza Jafari · Ching-Yi Lai
Academia Sinica · National Taiwan University · National Yang Ming Chiao Tung University

Why This Matters

QLDPC

Modern fault-tolerant code families need exact distance certification, but the solver landscape is still poorly understood.

Paper goal

Compare SAT, SMT, MaxSAT, and strong non-SAT baselines on a shared benchmark suite and identify what actually drives scalability.

Problem Setup
What are we solving?

Exact distance is a bottleneck

  • The code distance determines the minimum-weight logical operator.
  • That distance directly reflects worst-case error-correcting capability.
  • For modern QLDPC constructions, distance is often much harder to obtain than size or sparsity.
  • As code families scale, exact distance computation becomes a practical certification bottleneck.

CSS view used in the paper

HX HZᵀ = 0

dZ = min wt(z)
subject to HX zᵀ = 0 and z ∉ row(HZ)

dX = min wt(x)
subject to HZ xᵀ = 0 and x ∉ row(HX)

d = min(dX, dZ)

The hard part is not just finding a candidate logical operator, but proving there is no better one below a target weight.

Formulations
SAT vs MaxSAT

SAT formulation

Parity constraints

Encode stabilizer checks such as even-parity XOR relations.

Nontriviality constraints

Enforce that the candidate is not generated by stabilizer rows.

Weight bound

Use a cardinality constraint to ask whether wt(x) ≤ w.

Exact distance is found by solving a sequence of SAT instances for increasing thresholds.

Example

Suppose a row of HX touches qubits 1, 2, and 4, and we test whether there is a logical operator of weight at most 2.

parity: x₁ ⊕ x₂ ⊕ x₄ = 0
nontriviality: y₁ ∨ y₂
weight bound: x₁ + x₂ + x₃ + x₄ ≤ 2

query: “Is there any satisfying assignment?”
if SAT, then distance ≤ 2; if UNSAT, then distance > 2

MaxSAT formulation

Hard clauses

Parity and nontriviality constraints must always hold.

Soft clauses

Each variable gets a soft clause favoring 0, so the solver minimizes operator weight directly.

This avoids repeated threshold search and turns the task into an optimization problem.

Example

For four qubits, keep parity and nontriviality as hard clauses, then add one soft clause per variable:

hard: parity + nontriviality
soft: (¬x₁), (¬x₂), (¬x₃), (¬x₄)

assignment A: (1,0,0,0) satisfies 3 soft clauses
assignment B: (1,1,0,0) satisfies 2 soft clauses

MaxSAT prefers A, because it corresponds to a lower-weight logical operator.

Core question

  • Parity constraints are handled either natively as XORs or via Tseitin-style CNF translation.
  • Nontriviality is reduced to a dual-logical-operator test such as LXxT ≠ 0.
  • Cardinality constraints are where formulation choices become crucial: sequential-counter, totalizer, or binary.
Technical angle

The paper studies not only solver families, but also how the encoding interface changes solver behavior on the same QLDPC task.

Benchmarks
Formal benchmark table

Benchmark families used in the paper

Family Representative instances Count Row wt. Col wt. Structural character
BB / GB BB 72 12 6, BB 144 12 12, BB 288 12 ?, GB 144 12 8, GB 144 12 12 9 6-8 3-5 Algebraically regular, sparse circulant-style structure
LP LP 34 20 2, LP 238 44 6, LP 544 80 12, LP 1768 224 ? 8 8 5 Larger lifted-product search spaces with weaker regularity
TN TN 36 8 3, TN 72 8 8, TN 144 2 13, TN 250 10 15, TN 360 4 ? 10 6-12 4-9 Expander/Tanner style structure and difficult transition-region behavior

The naming convention is roughly {family} {n} {k} {d}, where ? denotes a distance not yet certified in the source benchmark collection.

Across all three families, the paper evaluates 27 total instances spanning small easy cases through hard transition-region benchmarks.

Evaluation
Study design

Experimental matrix

Axis What varies Role in the study
Benchmarks 27 QLDPC instances across BB/GB, LP, and TN families Expose easy, moderate, and transition-region behavior
Solvers 21 tools spanning SAT, SMT, MaxSAT, and non-SAT baselines Compare paradigms rather than one solver line
Configurations 46 representative configurations in the full heatmap Separate solver-family effects from encoding effects
Timeouts 3-minute pre-evaluation, 2-hour main evaluation Filter dominated encodings before large-scale comparison

Cardinality pre-evaluation

Encoding Solved Timeout Fastest
Sequential-counter 218 79 133
Totalizer 224 73 92
Binary 151 146 0
Implication

Binary encoding is empirically dominated, so the main experiments focus on sequential-counter and totalizer as the competitive formulations.

Finding 1
Optimization and cardinality dominate

XOR is not the whole story

The paper’s central claim is that the practical bottleneck is not parity reasoning alone. The harder part is handling cardinality constraints and proving optimality.

Interpretation

Even though the formulation is rich in XOR constraints, native XOR-aware SAT did not show a systematic advantage across the benchmark suite.

Why that matters

This challenges the intuition that “more XOR support” should automatically translate into better QLDPC distance solving.

What mattered more?

Optimization architecture
very high
Cardinality encoding
high
Native XOR support
limited

This visual is a presentation summary of the paper’s empirical conclusion, not a literal chart from the paper.

Finding 2
MaxSAT is the strongest frontier

Transition-region benchmarks tell the real story

Easy instances make many solvers look good. The informative cases are the transition-region benchmarks, where some methods still solve and others start timing out.

  • Branch-and-bound MaxSAT stayed competitive on the hardest cases.
  • MaxCDCL achieved the strongest overall coverage in the curated comparison.
  • Gurobi remained a serious baseline and was sometimes comparable or faster on hard instances.

Key separation

Stronger on hard cases

Branch-and-bound MaxSAT, MIP, and a few strong exact baselines.

Competitive but earlier plateau

Conventional SAT, SMT, XOR-aware SAT, and unsat-core-guided MaxSAT.

Important nuance

Not all MaxSAT solvers behave alike. Solver architecture itself is a major explanatory variable.

Finding 3
Old intuitions do not transfer cleanly

Brouwer-Zimmermann weakens in QLDPC

  • The Magma implementation often finds good candidate upper bounds.
  • But it is substantially weaker at proving that no better solution exists.
  • That means classical-code distance techniques do not carry over cleanly to the QLDPC regime.
Presentation takeaway

Finding a low-weight logical operator and certifying optimality are computationally different tasks.

Recommended hard benchmarks

  • BB 144 12 12
  • GB 144 12 12
  • BB 144 14 14
  • LP 442 68 10
  • LP 544 80 12
  • TN 144 2 13
  • TN 250 10 15

The paper argues these transition-region instances are the most discriminative for future solver comparisons.

One-Slide Summary
If you only remember four points

What this paper changes

1. QLDPC distance is not “just an XOR problem”

The hardest part is tied to cardinality management and optimality certification.

2. Optimization-oriented methods define the frontier

Especially branch-and-bound MaxSAT and strong MIP-based baselines.

3. MaxSAT is not one monolithic category

Branch-and-bound and unsat-core-guided approaches behave qualitatively differently.

4. Benchmark choice matters

Transition-region instances are far more informative than uniformly easy or hopeless ones.

Closing
Resources and next step

Suggested closing message

For exact QLDPC distance computation, solver success is driven less by parity reasoning than by how well the method handles optimization structure, cardinality constraints, and proof of optimality.

Reproducibility

  • The paper introduces an open experimental platform, DistQLDPC, derived from MaxCDCL.
  • It reports both certified lower and upper bounds when exact convergence is not reached.
  • This makes optimization-oriented runs easier to compare with incremental SAT workflows.

For a public-facing website version, we can later shorten the table-heavy slides and keep this fuller version as the seminar deck.