CayleyR: Solving the TopSpin puzzle via cycle intersection

Yuri Baramykov

REVIEW 4 major objections 5 minor 10 references

CayleyR: Solving the TopSpin puzzle via cycle intersection

T0 review · 4 major / 5 minor · reviewed 2026-08-02 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read This paper presents a randomized cycle-intersection algorithm that solves the TopSpin puzzle for state sizes up to n=20, returning verified paths in seconds.

desk verdict A real, working randomized TopSpin solver with verified paths, but the paper oversells scalability and contains a false group-theoretic claim. read the letter →

arxiv 2607.13219 v1 pith:AUIATSLN submitted 2026-07-14 cs.AI

Yuri Baramykov This is my paper · ORCID

classification cs.AI MSC 20B4005C2568W05

keywords TopSpinpuzzleCayleygraphcycleintersectionbidirectionalsearchrandomizedalgorithmpermutationgroupstatespacepathfinding

verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

The paper presents cayleyR, an R package that solves the TopSpin(n,k) permutation puzzle by searching for intersections between random cycles in the Cayley graph of the symmetric group. Instead of expanding states level by level, it generates random operation words, unrolls each word's periodic cycle from both start and target states, and looks for a shared permutation. When fronts miss, a distance heuristic picks bridge states and the cycle expansion restarts from them. The author claims this 'Iterative Cycle Intersection' algorithm reliably finds verified paths for n up to 20 in under a few seconds, without precomputed pattern databases. The significance is a practical solver for a puzzle with 20! states that does not require shortest-path guarantees.

What carries the argument

The central object is the cycle of an operation sequence (Definition 3): for a word w, the composed permutation φ_w has finite order, and its iterates σ0, φ_w(σ0), φ_w²(σ0), ... form a periodic cycle that returns to σ0. Recording all intermediate states during each application multiplies the coverage by the word length. The Iterative Cycle Intersection algorithm (Algorithm 1) alternates random cycle expansion from both fronts with distance-guided bridge selection, using a hash-indexed state store to find intersections in O(min(|S|,|F|)) time.

What would settle it

Run the solver on 100 random TopSpin(20,4) targets generated from 100-move scrambles, with the recommended parameter settings and a 30-second wall-clock limit per instance; if more than a small fraction fail to produce a verified path (or if the median time exceeds a few seconds), the central scalability claim collapses. Alternatively, engineer a single instance where the bridge-selection distance heuristic monotonically leads the fronts away from each other and observe non-termination.

Watch

Extended reading notes

Core claim

The core discovery is that a cycle of a random operation word—iterating a composed permutation φ_w until it returns to the seed—enumerates up to c·m distinct states for a word of length m and order c, so a modest number of cycles can cover a large region of the Cayley graph. Expanding these cycles simultaneously from the initial and final permutations and intersecting the two stores yields a connecting path; if no intersection appears, the algorithm re-roots the fronts at states that minimize a distance heuristic to the opposite endpoint. Every returned path is verified by re-applying the move sequence to the start state.

Load-bearing premise

The algorithm's termination rests on the unverified assumption that randomly sampled operation-word cycles, expanded from both fronts and re-rooted at heuristic bridge states, will intersect within at most T rounds; the paper provides no theoretical bound, and its own experiments show failures on 2–9 of 12 instances depending on the ranking strategy.

Editorial extensions

If this is right

TopSpin(20,4), with 20! states, becomes practically solvable on a desktop machine without precomputed pattern databases or IDA*.
Because the method returns verified paths (re-applying the sequence reproduces the target), it can be used as a reliable planner even when the path is not shortest.
The same machinery—random cycles, intersection, bridge selection—applies to any permutation puzzle whose generators are invertible elements of S_n, such as pancake sorting or Hungarian Rings.
The optional hub-based transport network (sparse BFS trees from each endpoint) reduces the effective distance before ICI starts, allowing the core search to close smaller gaps.
An optional depth-limited BFS post-processor can shorten many returned paths, sometimes by up to 50%, though savings vary by instance.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

The claimed reliability depends on an unproven stochastic-coverage premise: random cycles from opposite fronts must intersect within a bounded number of rounds. A natural extension would be to measure success-rate curves against state-space size and locate a phase transition where the method becomes unreliable.
The ranking of random words by 'most unique states' maximizes coverage but yields very long paths (up to ~74,000 moves in tests), whereas 'most repeated' is faster with short paths but less reliable; an adaptive strategy that switches ranking mid-search based on intersection progress might combine both strengths.
The celestial-coordinate embedding is currently only a visualization tool; if the holographic duality conjectured for Cayley graphs could be realized computationally, geometric intersection of lattice paths might replace hash lookups, potentially scaling to much larger n.
A testable extension is to apply the algorithm to a puzzle with a known difficult instance (e.g., the antipode of the TopSpin graph) and measure whether the distance heuristic leads bridge selection toward a dead end.

Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Desk Editor's Note

read the letter

Colleague,

You asked for a quick read on the cayleyR paper. The core is a genuinely new heuristic: instead of random walks or full BFS, it generates random operation words, expands their entire cycles from both endpoints, and looks for intersections, re-rooting at distance-based bridge states when they miss. That idea is not in the papers it cites, and the package actually runs: paths are reconstructed from visited states and re-verified by applying the move sequence, with code on CRAN. The hub-based BFS preprocessing that shrinks the effective gap before ICI is a sensible engineering addition, and the reported solve times on the instances shown (TopSpin(14,4) with varied scramble distance, plus n=10–16 with a 20-move scramble) are plausible. Table 2 shows a handful of sub-2-second solves with verified paths. Give credit for a concrete artifact.

The soft spots are real and the stress-test note lands. Definition 2's claim that the generators generate S_n for even k and A_n for odd k is wrong as stated: for k=1 the reversal is trivial (though k≥2 in the definition), and for k=n the generators produce the dihedral group, not S_n. The paper only experiments with k=4, so this doesn't invalidate the practical results, but a false theoretical foundation in a group-theory-flavored paper is a problem. More important, the 'n up to 20' claim in the introduction is not backed by experiments: Section 6.3 goes only to n=16, and those instances use a 20-move scramble—very easy. Table 1 does include n=14–30 but aggregates 12 instances across that range, with success rates from 25% to 83% depending on ranking strategy, and no per-size breakdown. So 'reliably finds paths in under a few seconds' is not established for n=20. The algorithm also has no termination guarantee; it rests on an unproven stochastic coverage assumption, and the high failure rates in Table 1 show that assumption is fragile. Finally, the evaluation lacks repeats, seeds, and baselines—no comparison against BFS, IDA*, or pattern databases on the same instances.

Minor: path lengths vary wildly (up to ~74,000 operations), the celestial-coordinates section is decorative at this stage, and the paper overreaches in claiming a fundamentally different strategy when it is really a stochastic meet-in-the-middle variant.

For a heuristic search audience this is a useful empirical contribution if the scaling claims are reined in. For a theory read, the missing guarantees and the group-theory error sink it. It deserves a serious referee—major revision, not desk reject.

Recommendation: send it to review, but the authors should fix Definition 2, re-run with seeds/repeats and proper baselines, and stop claiming n=20 until they have tested it.

Referee Report

4 major / 5 minor

Summary. The paper presents cayleyR, an R package (with a C++ backend and optional Vulkan GPU acceleration) for solving the TopSpin(n,k) permutation puzzle. The main algorithmic contribution is the Iterative Cycle Intersection (ICI) algorithm: from the start and target states, the algorithm samples random operation words, expands their cycles, stores visited states in a hash-indexed state store, and checks for intersection; when no intersection is found, it selects new bridge states using distance heuristics and repeats. The package also includes a sparse-BFS 'transport network' variant, path shortening, and a speculative 'celestial coordinates' visualization. The experiments report performance on TopSpin(14,4) with varying scramble distance, on TopSpin(10–16,4) with scramble distance 20, and a comparison of ranking criteria across 12 instances with n from 14 to 30. The paper claims scalability to n=20 and reliable solving in a few seconds.

Significance. The ICI idea is genuinely different from standard bidirectional BFS and pattern-database solvers: it replaces level-by-level exploration with algebraically generated cycles, and the package provides a public, reproducible implementation. A strength of the paper is that every returned path is verified by re-applying the move sequence, and the code is available on CRAN. If the n=20 scalability claim were supported, this would be a practically interesting solver for a 20!-state space. However, the manuscript as submitted does not support that claim: the experiments stop at n=16, and the n=15/n=16 cases are solved by distance-2 paths, so they are not meaningful scaling tests. Moreover, the reliability claim in Section 8 is contradicted by Table 1, where even the best ranking strategy fails on 2 of 12 instances and the worst fails on 9 of 12. The mathematical characterization in Definition 2 also has a boundary error. The core algorithmic idea remains plausible, but the paper needs substantial revision in its claims and theoretical framing before it can be accepted.

major comments (4)

[Section 1 vs. Section 6.3] Contribution (3) states the paper reports experiments 'demonstrating the scalability of the approach for TopSpin(n,k) instances with n up to 20.' Section 6.3 tests only n=10–16, and Section 6.2 explicitly says 'Experiments with GPU acceleration and scalability for n exceeding 20 will be reported in a future revision.' Moreover, in Table 3 the n=15 and n=16 rows have path length 2, meaning the random targets were at distance 2 from the identity; these instances are trivial and provide no evidence about scaling. The claim 'n up to 20' is therefore unsupported and should be either substantiated with experiments or removed.
[Section 8 and Table 1] The conclusion states the algorithm 'reliably finds paths in under a few seconds.' Table 1, however, reports success rates of 3/12 to 10/12 across the six ranking strategies, including 25% failure for 'most repeated' and 42% failure for 'longest.' Median times range from 0.0–161 s. Even the best strategy ('most unique') fails on 2 of 12 instances. Thus the reliability claim is not established. The paper should report per-instance outcomes, timeout behavior, and a confidence interval for success rates, or explicitly frame the method as a stochastic heuristic without a reliability guarantee.
[Definition 2, Eq. (1)–(3)] The statement that the TopSpin Cayley graph is Γ(S_n,{L,R,X_k}) with G=S_n if k is even and G=A_n if k is odd is false at the boundary k=n. For k=n, X_n is the full reversal, and the group generated by L, R, and X_n is the dihedral group of order 2n, not S_n or A_n. For example, TopSpin(4,4) has 8 states, not 24. The definition should be restricted to 2≤k<n and the exceptional cases stated. Since all experiments use k=4 with n≥10, this does not invalidate the empirical results, but the mathematical foundation as written is incorrect.
[Algorithm 1, Section 4.1] The algorithm's termination depends on the unproven stochastic-coverage assumption that random operation-word cycles expanded from two fronts will intersect within T rounds. There is no theoretical bound on the number of sampling rounds or on the probability of failure. Algorithm 1 can and does return FAIL (line 32), and Table 1 demonstrates failures in practice. For a randomized heuristic this is acceptable, but the paper should state explicitly that no guarantee is provided and should report the distribution of outcomes (including timeouts) rather than implying universal success. This is load-bearing because the central contribution is a practical solver, not a proof of existence.

minor comments (5)

[Definition 2] The condition '2≤k≤n' should be '2≤k<n' for the usual TopSpin puzzle; the k=n case is not only exceptional but also makes the parity claim false, as noted above. Please clarify the intended domain.
[Figures 1–2] The manuscript contains captions for Figures 1 and 2 but no actual images. Either include the figures or remove the captions.
[Sections 5.6 and 7.1] The 'celestial coordinates' feature and the holographic duality discussion are speculative and are not used in any experimental result. They should be moved to a clearly labeled future-work section or shortened, as they distract from the core algorithmic contribution.
[Table 1] The 'Typical path length' column reports extremely wide ranges (e.g., 4–73980). A range is not a good summary for such skewed data; report medians, quartiles, or per-instance values. The same applies to the time ranges.
[Section 7.1] Reference [5] is an unreviewed arXiv preprint. Its use as the basis for the holographic discussion should be explicitly labeled as speculative, and the claims should be softened accordingly.

Circularity Check

0 steps flagged · score 2.0 of 10

No circular derivation: paths are constructed from visited states and verified; mild self-referential benchmarking only.

full rationale

The ICI algorithm's output path is assembled from states actually visited during cycle expansion, and every returned path is checked by re-applying it to the start state (Section 5.3.3 Step 5, Example 9). No fitted parameter is renamed as a prediction: the random-combination ranking, bridge selection, and BFS/hub parameters are heuristics, not fit to the same data they are used to predict. The mathematical setup (Definition 2, Definition 3) is used to generate the search space, not to derive the solver's success; the group-theoretic claim about even k generating S_n is a stated premise, not a result derived from the algorithm. Self-citations ([2],[3]) are package availability references and are not load-bearing for the central method. The paper's own Tables show stochastic failures and unproven termination, and Section 6.2 explicitly defers n>20 and GPU experiments to a future revision, which weakens the scalability claims but is a correctness/evidence concern, not circularity. The only mild self-referentiality is that targets are produced by the same package's random walk and success is judged by its own verifier, with no external optimal-distance benchmark; this does not make the derivation circular.

Assumptions & free parameters 7 free parameters · 4 assumptions · 1 invented entities

The central claim rests on heuristic search rather than a theorem. The algorithm's success depends on several hand-set hyperparameters (listed above) and on unproven assumptions about random-cycle coverage and distance heuristics; the only invented entity is the celestial-coordinate embedding, which is decorative and explicitly not used for the reported results.

free parameters (7)

combo_length = 25
Length of random operation words; chosen by hand for all experiments (Section 6.2); no sensitivity analysis.
n_samples = 400
Number of random words sampled per round; results depend on this sampling budget.
n_top = 100
Number of words retained for full cycle expansion after ranking; controls store size and coverage.
max_iterations = 150
Termination cap for ICI rounds; if coverage is poor the algorithm simply fails.
potc = 1
Fraction of cycle states retained in the store; 1 means full cycle expansion, affecting memory and intersection chances.
ptr = 3
Limits number of intersection candidates reconstructed per round; affects path quality and runtime.
bfs_hub_parameters = levels=200, hubs=7, random=3
Sparse BFS tree parameters in find_path_bfs (Section 5.3.3); chosen by hand and directly affect hub distance and path quality.

assumptions (4)

ad hoc to paper TopSpin state space is the Cayley graph of S_n for even k and A_n for odd k.
Stated in Definition 2 and Section 1 without proof; false for (n,k) pairs such as k=4 with odd n, where both L and X_4 are even permutations, so the generated group is A_n, not S_n.
domain assumption Random cycles from sampled words cover enough of the state space that the two fronts intersect before T rounds.
Algorithm 1 relies on this; no theoretical bound is given, and Table 1 failure rates for several ranking strategies contradict it.
domain assumption Manhattan and breakpoint distances are useful proxies for Cayley-graph distance in bridge selection.
Used in Algorithm 1 phase 4 and Section 5.3.3; no evidence of correlation with true Cayley-graph distance is provided.
domain assumption Randomly generated targets are representative difficult instances.
Targets are generated by random walks (generate_state); several are trivially close to the identity (path length 2), so scaling results partly reflect easy instances.

invented entities (1)

Celestial coordinates
purpose: Embed search history on the unit sphere for visualization; proposed as an experimental distance for bridge selection.
Section 5.6 maps cumulative move counts (nL,nR,nX) via stereographic projection. The authors state it does not capture graph distance and is not more effective than Manhattan/breakpoint, and it plays no role in the reported results.

how reviews work

0 comments

Cite this review

Pith. "Pith review of CayleyR: Solving the TopSpin puzzle via cycle intersection." pith.science (2026). https://pith.science/paper/AUIATSLN

@misc{pith2026260713219,
  author       = {Pith},
  title        = {Pith review of: CayleyR: Solving the TopSpin puzzle via cycle intersection},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/AUIATSLN}},
  note         = {Machine review of arXiv:2607.13219}
}

read the original abstract

We present cayleyR, an R package for solving permutation puzzles by detecting cycle intersections in Cayley graphs. The core algorithm performs an iterative bidirectional search: from both the initial and target permutation states, random operation sequences generate cycles in the Cayley graph of the symmetric group Sn; their intersection yields a connecting path. When no direct intersection is found, a distance-guided bridge selection narrows the gap, and the process repeats. The package targets the TopSpin(n,k) puzzle, whose state space is a Cayley graph of Sn generated by a cyclic shift and a prefix reversal. We describe the mathematical framework, the algorithm, and its implementation, which combines a C++ hash-indexed state store with optional Vulkan GPU acceleration. The software is publicly available on CRAN.

Figures

Figures reproduced from arXiv: 2607.13219 by the authors.

**Figure 1.** Three-dimensional layout of cycle states for TopSpin(7, 4) using celestial coordinates (θ, ϕ, ω). Light blue points represent all states visited during cycle analysis of multiple operation sequences; edges connect consecutive states within each cycle. The red overlay highlights the longest cycle found. Visualization produced by the cgvR package (in development). Several concrete extensions are planned for future re… view at source ↗

**Figure 2.** Celestial coordinate embedding of cycle expansion states for TopSpin(20, 4). Each point is a permutation state mapped to S 2 via stereographic projection of cumulative operation counts (nL, nR, nX); colors encode different operation sequences. The red chain traces the recovered path through successive bridge states. Spatial clustering reflects that similar operation sequences produce nearby celestial positions. pa… view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

10 extracted references · 2 canonical work pages

[1]

Bortoluzzi, P.P. (2012). A pattern database approach for solving the TopSpin puzzle problem.Bachelor thesis, University of Basel

2012
[2]

(2026).cayleyR: Cayley Graph Analysis for Permutation Puzzles

Baramykov, Y. (2026).cayleyR: Cayley Graph Analysis for Permutation Puzzles. R package version 0.2.1, CRAN.https://CRAN.R-project.org/package=cayleyR(DOI: 10.32614/CRAN.package.cayleyR)

work page doi:10.32614/cran.package.cayleyr 2026
[3]

(2026).cgvR: Interactive 3D Visualization of Large Cayley Graphs via Vulkan

Baramykov, Y. (2026).cgvR: Interactive 3D Visualization of Large Cayley Graphs via Vulkan. R package version 0.1.2, CRAN.https://CRAN.R-project.org/package=cgvR(DOI: 10.32614/CRAN.package.cgvR)

work page doi:10.32614/cran.package.cgvr 2026
[4]

Chervov, A., Soibelman, A. (2025). CayleyPy RL: Pathfinding and reinforcement learning on Cayley graphs. arXiv preprintarXiv:2502.18663

arXiv 2025
[5]

Chervov, A., Levkovich-Maslyuk, F., Smolensky, A., et al. (2026). CayleyPy-4: AI-Holography. Towards analogs of holographic string dualities for AI tasks.arXiv preprintarXiv:2603.22195

arXiv 2026
[6]

(1988).Group Representations in Probability and Statistics

Diaconis, P. (1988).Group Representations in Probability and Statistics. IMS Lecture Notes, Vol. 11

1988
[7]

Goldstein, R. (2015). The group theory of TopSpin.Senior thesis, Bard College

2015
[8]

Pasterski, S., Pate, M., Raclariu, A.-M. (2021). Celestial holography.arXiv preprintarXiv:2111.11392

arXiv 2021

Show all 10 references

[9]

Ratner, D., Warmuth, M.K. (1990). The (n 2 −1)-puzzle and related relocation problems.J. Symbolic Comput., 10(2), 111–137

1990
[10]

Senac, E. (2016). The 15 puzzle and TopSpin.Butler University Undergraduate Research Conference. Address Email address:lbsbmsu@mail.ru

2016

Pith tools

Reviewed August 2, 2026 · model on record in the stance chip above.

[1] [1]

Bortoluzzi, P.P. (2012). A pattern database approach for solving the TopSpin puzzle problem.Bachelor thesis, University of Basel

2012

[2] [2]

(2026).cayleyR: Cayley Graph Analysis for Permutation Puzzles

Baramykov, Y. (2026).cayleyR: Cayley Graph Analysis for Permutation Puzzles. R package version 0.2.1, CRAN.https://CRAN.R-project.org/package=cayleyR(DOI: 10.32614/CRAN.package.cayleyR)

work page doi:10.32614/cran.package.cayleyr 2026

[3] [3]

(2026).cgvR: Interactive 3D Visualization of Large Cayley Graphs via Vulkan

Baramykov, Y. (2026).cgvR: Interactive 3D Visualization of Large Cayley Graphs via Vulkan. R package version 0.1.2, CRAN.https://CRAN.R-project.org/package=cgvR(DOI: 10.32614/CRAN.package.cgvR)

work page doi:10.32614/cran.package.cgvr 2026

[4] [4]

Chervov, A., Soibelman, A. (2025). CayleyPy RL: Pathfinding and reinforcement learning on Cayley graphs. arXiv preprintarXiv:2502.18663

arXiv 2025

[5] [5]

Chervov, A., Levkovich-Maslyuk, F., Smolensky, A., et al. (2026). CayleyPy-4: AI-Holography. Towards analogs of holographic string dualities for AI tasks.arXiv preprintarXiv:2603.22195

arXiv 2026

[6] [6]

(1988).Group Representations in Probability and Statistics

Diaconis, P. (1988).Group Representations in Probability and Statistics. IMS Lecture Notes, Vol. 11

1988

[7] [7]

Goldstein, R. (2015). The group theory of TopSpin.Senior thesis, Bard College

2015

[8] [8]

Pasterski, S., Pate, M., Raclariu, A.-M. (2021). Celestial holography.arXiv preprintarXiv:2111.11392

arXiv 2021