{"id":"a06bbbf2-6d67-44fa-924f-f08e3c969663","arxiv_id":"2608.00002","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":1,"one_line_summary":"Under lattice-direction and planar-edge conditions, exactly 26 soft, space-filling cubes and 68 fundamental domains exist.","lead":"This paper classifies all ways to round the corners of cubes so that the resulting shapes still fill space without gaps, under the rules that edges follow lattice directions and lie in planes. It reports exactly 26 distinct 'soft cube' shapes and 68 periodic building blocks, and ships a short Python enumeration.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Theorem 1's proof assumes, without justification, that every complete set of softening equations has the form of a bijection between A and B plus one intra-subset equation; if other complete sets exist, the 'exactly 26' classification could miss valid soft cells.","rationale":"The reader's weakest assumption points to the WLOG labeling split in Theorem 1. I agree that Theorem 1 is the foundation of the 'exactly 26' claim and that its proof has a serious gap. However, I would locate the core issue slightly differently: the A/B split itself is likely salvageable by relabeling the six half-tangents so that the complementary pairs are (a,c) and (b,d). The truly unproved step is the structural restriction that complete sets of softening equations must consist of a bijection between A and B plus one additional intra-A equation. This restriction is asserted without proof and is not a consequence of the split. A concrete counterexample to the restriction exists at the node level: the four cross equations (a,b),(b,c),(c,d),(d,a) make all eight corners soft under a consistent collinear assignment, yet this set is absent from Table 1. Whether such a node configuration can be part of a space-filling tiling is not addressed in Theorem 1; it is deferred to Theorem 3, whose proof is also sketchy. Thus the central 'exactly 26' claim is not yet established. The reader's verdict of CONDITIONAL remains appropriate: the paper presents a plausible and novel classification with a reproducible Python code, but a finite computational check of Theorem 1's exhaustiveness would either confirm the 26 count or reveal missing cases. I agree with the reader that this is addressable and that the verdict should not change until such a check is performed.","tokens_in":14823,"tokens_out":36806,"duration_ms":359275,"concrete_test":"Brute-force all assignments of the six labels {a,...,f} to the six lattice directions {±x,±y,±z} (or, equivalently, enumerate all subsets of the 12 face-pair equations up to size 6) and retain those in which every corner triple eab,...,fda contains at least one opposite pair, using the transitive implications of the equations. Classify the resulting nodal configurations up to cube symmetries and relabeling, and compare with the 12 equation systems of Table 1. If any configuration not equivalent to one of the 12 emerges—especially one that also satisfies the prismatic condition (at least one straight edge, three non-coincident lines)—then Theorem 1 is incomplete and the 26 count must be re-derived. This finite test can be implemented by extending the paper's own Python code to enumerate equation sets rather than just face pairings.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The most load-bearing gap is in Section 2, Theorem 1, immediately after Eq. (2). The proof asserts 'The general case is obtained by specifying three pairs of edges whose half-tangents, after softening, become complementary half-lines' and then restricts complete sets of softening equations to a bijection C_{i,j} ⊂ A_i×B_i (three cross pairs covering all six labels) plus a single additional equation from A_i, with only (a,e) and (c,e) considered (Table 1). No argument shows that every set of dot=-1 constraints satisfying the lattice-direction and planar-edge conditions must have this form. The A/B split may be WLOG up to relabeling, but the equation-form restriction is independent and unproved. For example, the four cross pairs (a,b),(b,c),(c,d),(d,a) form a perfect matching of the corner graph and, when solved, yield a consistent assignment (all labels on one line, alternating signs) in which all eight corners are soft; this set is not among the twelve in Table 1. Whether such degenerate configurations can be extended to space-filling tilings is not settled in Theorem 1; it is deferred to Theorem 3. Since Theorem 4's 'exactly 26' depends on the exhaustiveness of the 12 systems, an incomplete enumeration would invalidate the central claim. The proof also relies on 'it can be verified' in Theorem 2, and Algorithm 2 does not itself test gap-free/overlap-free tiling, compounding the issue.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper classifies soft tilings of the cubic lattice under the restrictions that edge half-tangents point along lattice directions and that edges are planar. The central claim is that, up to the equivalence defined in Theorem 4, there are exactly 26 parallel soft tilings that partially agree to second order with the cubic lattice, and that these give rise to 68 fundamental domains in 8 lattice-symmetry classes. The argument proceeds by (i) enumerating complete sets of softening equations at a node (Theorem 1), (ii) selecting the node configuration that softens all eight corners (Theorem 2), (iii) proving that any gap-free, overlap-free soft tiling must be prismatic (Theorem 3), (iv) classifying curved edge and face types (Lemma 1), and (v) enumerating cells from face pairings via an algorithm with a Python implementation (Theorem 4). A symmetry classification and fundamental-domain counts are then presented (Theorem 5).","tokens_in":15230,"tokens_out":13257,"duration_ms":149312,"significance":"If the central classification is correct, the paper provides a valuable and nontrivial complete enumeration: it goes beyond existence statements for soft cells and gives a concrete list of geometries under natural constraints, together with explicit symmetry data. The manuscript includes several strengths that should be acknowledged: it ships the Python code used for the enumeration, it provides extensive tables and figures for all 26 cells and 68 fundamental domains, and it makes precise, falsifiable numerical claims (26, 68, 8). These strengths make the paper potentially useful as a reference classification. However, the exactness of the counts is not currently supported to the standard required by a journal proof: the load-bearing enumeration in Theorem 1 is asserted rather than proved, and the reproduced code does not implement the equivalence relation used in the main theorem. The result may well be correct, but the manuscript needs additional rigorous support before the classification can be accepted.","major_comments":[{"comment":"The proof asserts, without demonstration, that every complete set of softening equations has the form of a bijection C_i,j between A_i={a,c,e} and B_i={b,d,f}, plus one intra-subset equation (a,e) or (c,e). The definition of a complete set only requires at least one equation in every vertex set; the four-equation lower bound does not force this particular structural form. The proof rules out neither non-bijective sets of cross-pairs nor a fourth equation lying in B_i or being another cross-pair, after the complementary pairs ac and bd are imposed. The 'general case is obtained by specifying three pairs...' sentence is an assumption, not a derivation. Since the 12 systems of Table 1 are the sole input to Theorem 4's 'exactly 26', this gap is load-bearing. A rigorous exhaustive enumeration, or a machine-checkable certificate that all complete sets reducing to these 12 systems, is needed.","section":"Section 2, Theorem 1 proof, after Eq. (2)"},{"comment":"The proof of Theorem 2 relies on 'It can be verified' to assert that exactly one of the 12 systems, in four equivalent guises, softens all eight corners. This is a finite but nontrivial check: one has to solve each of the 12 systems under the lattice-direction restriction, list the resulting half-tangent assignments, and verify the corner-softening condition for each of the eight corners. The claim that 1A and 1B fail in the stated corners is plausible but not shown. Since Theorem 2 is what identifies the unique admissible node configuration used in all later steps, the verification should be written out or delegated to a small reproducible script included with the paper.","section":"Section 2, Theorem 2 proof"},{"comment":"Lemma 1 states that the admissible node configuration yields exactly four curved edge types and six curved face types, with the sentence 'More curved face types cannot be realized while preserving the softness conditions.' No derivation of this exhaustiveness is given; the classification is justified by schematic inspection of Figures 6--8. The later 26-cell count depends directly on this lemma. Please provide either a proof from the endpoint half-tangent conditions that no other edge and face types are possible, or an independent exhaustive enumeration of face types that is shown to be complete.","section":"Section 3, Lemma 1"},{"comment":"Algorithm 2, as reproduced in the Appendix, does not implement the equivalence relation stated in Theorem 4. The text defines two cells as identical if they are related by rotations and/or reflection, and Algorithm 1 uses that definition. The Python code, however, defines transformations only as z-rotations, y-rotations, and their compositions; there is no mirror transformation. The paper itself observes that without mirror transformations the output is 43 copies (9 achiral + 17 chiral multiplied by 2), not 26. Thus the code, as printed, supports the 43-count and not the 26-count of Theorem 4. The exactness of 26 currently rests on the manual Algorithm 1, not on the reproducible algorithm. Moreover, neither Algorithm 2 nor the code explicitly checks that the paired cells tile space without gaps or overlaps; it relies on Theorem 3 without verifying that the resulting pairings satisfy the","section":"Section 4, Theorem 4 proof and Appendix code"}],"minor_comments":[{"comment":"The phrase 'polihedric tiling' appears to be a typo for 'polyhedric tiling'.","section":"Definition 1"},{"comment":"The numbering of the steps is confusing: Step 4* is listed after Step 4 but used instead of Step 4; consider renumbering or denoting it as a variant clearly.","section":"Section 1.1, EEB algorithm"},{"comment":"The symmetry-group illustrations point to an external website (newton.ex.ac.uk). This is not a stable reference; please include the group names in text and cite a permanent source.","section":"Figures 13--21"},{"comment":"The notation for the subsets is inconsistent in places: A_i and B_i are used in the text and equations, while Table 1 uses A and B only. This should be harmonized for readability.","section":"Table 1 and notation"},{"comment":"The code would benefit from comments explaining that it deliberately omits mirror transformations, and from an explicit output showing whether the printed length is 43 or 26; currently the text says 'The algorithm shows 26' while the code description says 43 without mirror transformations.","section":"Appendix Python code"}],"recommendation":"major_revision","confidential_remarks":null},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The paper gives a novel and plausible census: 26 soft cubic cells and 68 fundamental domains for the cubic lattice under natural conditions, extending the group's earlier work on truncated octahedra. The figures, tables, and the short Python program are genuinely useful, and the prismatic-structure result (Theorem 3) is a nice step that reduces the problem to pairing curved faces. If the enumeration is right, this is a solid contribution to the soft-tiling niche.\n\nThe soft spot is in the middle. Theorem 1's proof asserts, without real justification, that every complete set of softening equations has the form of three cross-pairs (a bijection between A and B) plus one intra-subset equation. That is not a harmless WLOG. The stress-test example—(a,b),(b,c),(c,d),(d,a)—covers all eight corners and is not in Table 1. I suspect it is degenerate and unlikely to extend to a space-filling tiling, but the proof does not say why, and Theorem 4's 'exactly 26' rests on the exhaustiveness of these 12 systems. Similarly, Theorem 2 leans on 'it can be verified', and Algorithm 2 classifies face pairs but does not itself test gap-free/overlap-free tiling; that property is carried by Theorem 3 and the construction, but the gap between the algorithm's output and the tiling claim is not fully bridged. These are addressable, but they are load-bearing, not cosmetic.\n\nThe self-citation pattern is not a real problem here: the prior framework is the natural starting point, and the new work extends it. What is missing is a formal enumeration argument, an explicit check that all 26 cells actually tile, and an independent verification of the six face types in Lemma 1.\n\nWho is this for? Researchers in soft cells, monohedral tilings, or geometric modeling of cellular materials. It deserves a serious referee—this is a real classification attempt with visible evidence, not a toy—but it should not be accepted as is. A revision that turns Theorem 1 into a real proof and makes the tiling checks explicit would move it from 'conditional' to 'accept'. I would not cite it in my own work until those gaps are closed.","headline":"A plausible classification of 26 soft cubic cells, but the central 'exactly' claim leans on an unjustified WLOG in Theorem 1 that needs fixing before publication.","tokens_in":15646,"tokens_out":3811,"would_cite":false,"duration_ms":41455,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["52C20","52C22","05B45"],"pacs":[],"model":"deepseek-v4-flash","headline":"Exactly 26 geometrically distinct soft cube cells exist that partially agree with the cubic lattice to second order under lattice-direction and planar-edge constraints, and they generate 68 fundamental domains in 8 symmetry classes.","keywords":["soft cells","space-filling tilings","cubic lattice","monohedral tilings","soft cubes","prismatic tilings","fundamental domains","chiral tilings"],"falsifier":"Run an exhaustive computer search over all assignments of six unit vectors drawn from the six cubic lattice directions to the six edge half-tangents at a node, imposing only the softness condition that each of the eight corner triples contains an opposite pair, and do not assume the A_i/B_i split used in Theorem 1. If any node configuration outside the 12 equation sets of Table 1 is found, the 26-cell classification is incomplete; if none is found, the missing-proof concern is settled.","tokens_in":14666,"feed_emoji":"🧊","tokens_out":5592,"duration_ms":52915,"temperature":0.7,"pith_summary":"The paper sets out to answer a question left open by earlier work: when the standard cubic lattice tiling is softened so that no tile has sharp corners, how many essentially different soft cubes are there? It proves that, under two natural restrictions — edge half-tangents must point in lattice directions and tile edges must be planar — there are exactly 26 geometrically distinct soft cube cells, all of them prismatic. This is a complete classification: any soft tiling meeting those conditions must be one of the 26. The paper further shows that the 26 cells generate 68 distinct minimal translational units, or fundamental domains, which fall into 8 lattice-symmetry classes, and that none of the tilings retains the full symmetry of the cubic lattice. A reader should care because the classification turns an open-ended existence question into a finite list, with an algorithmic method that could be reused for other polyhedral tilings.","feed_headline":"Exactly 26 soft cube shapes can tile space","feed_subtitle":"Cubic-lattice softenings are all prismatic and form 68 fundamental domains in 8 symmetry classes.","key_machinery":"The load-bearing object is the 'softening equation' u_i · u_j = −1, expressing that two edge half-tangents at a node are opposite, so the two edges join into a smooth curve. The Extended Edge Bending algorithm, run with the identity symmetry group and with half-tangents restricted to the six cubic lattice directions, converts the local softness condition into 12 complete sets of such equations; the admissible node configuration from Theorem 2 is then propagated through a cell by rotating it at each of the eight vertices. The combinatorial catalogue is organized by four curved edge types (1,1',2,2', depending on whether endpoint tangents coincide or are opposite) and the six curved face types","core_discovery":"The central claim is Theorem 4: up to rotation, reflection, and composition, there are exactly 26 parallel soft tilings that partially agree to second order with the cubic lattice. Here 'partially agree to second order' means the node positions are those of the cubic lattice and the half-tangent unit vectors of the softened edges are a subset of the six original lattice directions; 'parallel' adds that every edge is a planar curve. The proof route is Theorem 1: the softening equations admit exactly 12 complete solution sets; Theorem 2: only one of these (up to equivalence) makes all eight corners at a node soft; Theorem 3: any gap-free soft tiling built this way is prismatic, i.e. it has a p","pith_inferences":["Beyond the paper: if the assumed node splitting A_i={a,c,e}, B_i={b,d,f} were relaxed, additional node configurations might exist; a brute-force enumeration over all ways to choose six lattice-direction unit vectors at a node, without that split, would test whether 26 is truly exhaustive.","Beyond the paper: the prismatic character suggests the 26 soft cubes are, in a precise sense, extruded versions of 2D soft square tilings; studying the 2D analogue might yield an independent derivation of the list.","Beyond the paper: the method of fixing lattice directions and planarity could be applied to other space-filling polyhedra, such as truncated octahedra or tetrahedral frameworks, to produce analogous finite classifications with crystallographic relevance.","Beyond the paper: the 8 lattice-symmetry classes of fundamental domains may correspond to specific crystallographic space groups, which could be used to search for photonic or phononic crystals built from soft cells."],"forward_implications":["If the classification is right, the search for soft cubic cells is closed: every lattice-directed, planar-edged soft cube is on the 26-cell list, so no further examples exist under those conditions.","Any such soft tiling is prismatic, so the full 3D tiling can be understood by studying a stack of square-lattice layers, reducing the problem to 2D.","The 68 fundamental domains, sorted into 8 lattice-symmetry classes, provide a catalog of minimal translational building blocks for space-filling soft shapes.","The enumeration shows that no soft cube tiling keeps the full cubic symmetry; the largest symmetry groups have order 8, which limits how symmetric a soft cubic tiling can be.","The accompanying Python algorithm gives a concrete way to regenerate the 26 cells and to check whether a proposed cell tiles achirally or requires mirror copies."],"fun_headline_variants":["All soft cubic tilings: 26 cells, 68 domains","Soft cubes: 26 geometries, 8 symmetries","Prismatic soft cubes solved: 26 types","Count them all: 26 soft cube cells","Exactly 26 ways to soften a cube"],"cache_read_input_tokens":2304,"weakest_assumption_plain":"The enumeration hinges on the unproved assumption, made in the proof of Theorem 1 (Section 2, just after Eq. (2)), that the six half-tangents at a node can always be relabeled as A_i={a,c,e} and B_i={b,d,f} with ac and bd already complementary; if a valid soft node cannot be represented this way, the 'exactly 26' count could miss cases.","fun_headline_variants_meta":{"raw":{"variants":["All soft cubic tilings: 26 cells, 68 domains","Soft cubes: 26 geometries, 8 symmetries","Prismatic soft cubes solved: 26 types","Count them all: 26 soft cube cells","Exactly 26 ways to soften a cube"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000198,"raw_usage":{"total_tokens":1228,"prompt_tokens":788,"completion_tokens":440,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":532,"completion_tokens_details":{"reasoning_tokens":364}},"tokens_in":532,"tokens_out":440,"duration_ms":5359,"temperature":1.0,"reasoning_tokens":364,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-04T05:08:41.287498+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Run an exhaustive computer search over all assignments of six unit vectors drawn from the six cubic lattice directions to the six edge half-tangents at a node, imposing only the softness condition that each of the eight corner triples contains an opposite pair, and do not assume the A_i/B_i split used in Theorem 1. If any node configuration outside the 12 equation sets of Table 1 is found, the 26-cell classification is incomplete; if none is found, the missing-proof concern is settled.","supporting_citations":[],"review_version":1}