{"id":"9ef12623-fc5c-4b21-acf5-2e1f5c2d253b","arxiv_id":"2411.14087","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"For generalized Zetterberg codes over odd-characteristic fields with q^s ≡ 7 mod 8, the covering radius is 2 for s=1, 3 for s≥s*, and in general characterized by a solvability condition; twisted half versions inherit these values.","lead":"This paper solves an open case in coding theory by determining the covering radius, the worst-case number of errors a code can be guaranteed to handle, for a family of generalized Zetterberg codes over odd-characteristic finite fields. The result completes the picture for these codes and yields new quasi-perfect codes, which are useful in error correction and data compression.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Example 4.15 explicitly leaves undetermined whether 7 and 9 lie in I(47), so the covering radii of C_7(47) and C_9(47) are not known; the paper's unqualified claim to have determined all generalized Zetterberg codes is therefore not established as stated.","rationale":"The reader's weakest_assumption was the genus formula quoted from [31, Theorem IV.1], which is a legitimate auxiliary assumption but one that is referenced to a published source. The most load-bearing concern about the central claim is more direct: the paper itself, in Example 4.15, leaves the covering radius of C_7(47) and C_9(47) undetermined, and the abstract's 'all generalized Zetterberg codes' is therefore not literally established. The reader did note this in the rationale ('the phrase all generalized Zetterberg codes should be read with care'), but did not elevate it to the weakest assumption. My read agrees with the reader's CONDITIONAL verdict: the mathematical machinery in Sections 3 and 4 is substantial and plausibly correct, but the unqualified completeness claim needs either more computation or a precise qualification. I did not choose Theorem 4.2's omitted proof as the primary concern because it is presented as a routine adaptation of [31] and may be acceptable in a journal context, although it deserves attention; the unresolved cases are a concrete, self-admitted gap in the central claim. A targeted test that decides the two missing memberships of I(47), or a revised claim that states the actual scope, would resolve this concern without changing the overall CONDITIONAL status.","tokens_in":21439,"tokens_out":16559,"duration_ms":155676,"concrete_test":"Determine whether 7 and 9 belong to I(47) by deciding, for q0 = 47 and s = 7 and s = 9, whether there exists an index i in {0,...,15} such that the system in Theorem 4.5 (for even i) or Theorem 4.6 (for odd i) has a solution with x, y1,...,ym ∈ F_{47^s}^*. A positive or negative decision for both s-values settles whether the 'all codes' claim can stand; if the direct computation is infeasible, the authors should either provide a sharper argument or explicitly restrict the main theorem to the cases that are actually determined.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim, repeated in the abstract, Corollary 3.14, Theorem 4.2, and the reader's strongest_claim, is that the covering radius of every generalized Zetterberg code in odd characteristic is now determined. Example 4.15 contradicts this: it states that the membership of 7 and 9 in I(47) is unknown, and therefore I(47) is only known to be {s ≥ 11 : s odd} ∪ J with J ⊆ {7, 9}. Since I(q0) is defined as the set of odd s with ρ(C_s(q0)) = 3, not knowing whether 7 ∈ I(47) means not knowing whether ρ(C_7(47)) is 2 or 3; likewise for s = 9. Moreover, for arbitrary q0 and odd s below the threshold s* of Eq. (45), no general decision is made: Theorem 4.2 gives a criterion in terms of Property NP_i, but the criterion is not evaluated for these cases, and Examples 4.12-4.15 cover only the smallest q0. Thus the unqualified statement that all codes have known covering radius is not supported. The issue is a stated limitation, not a hidden error, and it can be repaired either by completing the two unresolved cases or by weakening the claim to 'all codes except possibly a finite list' or 'determined up to a finite computation'.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper studies generalized Zetterberg codes C_s(q0) over finite fields of odd characteristic in the remaining congruence class q0^s ≡ 7 (mod 8). The authors prove that the covering radius of such a code is at most 3, give an exact dichotomy (radius 3 if a certain Property NP_i holds and 2 otherwise), show that the radius is 2 for s = 1, and prove that for odd s beyond a threshold s* defined in Eq. (45) the radius is 3. They also define twisted half generalized Zetterberg codes and show that these codes inherit the same covering radii, yielding quasi-perfect codes for some parameters. The main techniques are elementary finite-field arithmetic, decomposition of F_{q^2} into cosets, Weil sums, and a Hasse-Weil estimate on a fiber product of curves.","tokens_in":21740,"tokens_out":18295,"duration_ms":170120,"significance":"If the arguments are completed and the overclaim is corrected, this is a substantial contribution: it would settle the open problem from [31] for the remaining congruence class, up to a small number of explicit undecided cases, and it would provide the first exact covering-radius results for this family in the hard case. The paper contains genuinely self-contained parts, especially the upper bound ρ ≤ 3 in Section 3, and it gives a concrete system-of-equations criterion for the exact radius. The introduction of twisted half generalized Zetterberg codes is a useful by-product and yields new quasi-perfect codes. However, the paper as written claims more than it proves: its own examples leave some covering radii undetermined, and a few load-bearing proof steps are skipped or quoted from [31].","major_comments":[{"comment":"The paper's central claim that the covering radius of every generalized Zetterberg code in odd characteristic is now determined is not supported by the manuscript's own examples. Example 4.15 explicitly states that membership of 7 and 9 in I(47) is unknown, so the covering radii of C_7(47) and C_9(47) are not determined; Example 4.14 leaves the status of 7 in I(31) open. Since I(q0) is defined as the set of odd s with ρ(C_s(q0)) = 3, and Theorem 4.2 only gives a criterion that is not evaluated for these small s, the abstract's claim and the sentence in Section 4 that 'We solve this open problem completely' overstate the result. Please either settle these finitely many undecided cases or revise the claims throughout to say that the covering radius is determined for all codes except possibly an explicit finite list, or 'up to a finite computation'.","section":"Abstract; Section 4, Examples 4.14–4.15"},{"comment":"Theorem 4.2 is the central characterization of when the covering radius equals 3, yet its proof is skipped with the remark that it uses the same arguments as [31]. Because [31] explicitly left the case q0^s ≡ 7 (mod 8) open, this is not literally a theorem proved in [31], and the step from the upper bound ρ ≤ 3 in Corollary 3.14 to the exact dichotomy 'ρ = 3 if and only if some Property NP_i holds' is not immediate. Please supply a proof or a detailed reduction, in particular showing how the results of Section 3 and the properties of H_m combine to rule out ρ ≤ 2 when no NP_i holds.","section":"Theorem 4.2"},{"comment":"The Weil bound for the sum of η((1 − x^2)Δ(x)) is asserted without proving that the degree-6 polynomial has no square factor; Proposition 3.8 only establishes that Δ(x) itself is not a square. If (1 − x^2)Δ(x) were a non-square constant times a square, the sum would be of size ≫ q rather than O(√q), and the positivity estimate N1 > 0 for q > 94 would be invalid. This is repairable: one can show Δ(±1) = 0 and that the remaining quadratic factor has discriminant 4(α^2 + 1)/α^2 ≠ 0, so the polynomial is not a constant times a square, but the argument should be included in the paper.","section":"Theorem 3.12, Eq. (36)"},{"comment":"The Hasse-Weil lower bound in Eq. (48) uses the genus of the fiber product χ, quoted as 1 + 2^{m−1}(m − 2) from 'the proof of [31, Theorem IV.1]'. This genus value is load-bearing for the estimate N(χ;s) > 2^m and hence for the threshold s* in Eq. (45), but it is not recomputed and no precise theorem statement is cited. Please either reproduce the genus computation or cite the exact numbered statement in [31] where this formula is proved, and confirm that it applies to the curve as defined here for all m.","section":"Theorem 4.8, Eq. (48)"}],"minor_comments":[{"comment":"The list of small q for which a direct Magma check is used includes 87, but 87 is not a prime power and no field F_87 exists; the valid set is {7, 23, 31, 47, 71, 79}.","section":"Theorem 3.12"},{"comment":"The Magma computations are described only in words, and some are said to be 'extremely time consuming' or to have been stopped. Please provide the Magma scripts or enough computational details to make the finite-field checks reproducible.","section":"Theorems 3.12–3.13 and Examples 4.12–4.15"},{"comment":"There are several typos: 'Theoprem' for 'Theorem' in the citations in Theorem 3.12, 'D1 and D2 are are nonzero squares' in Proposition 4.3, 'Zettenberg' for 'Zetterberg' in Definition 5.1, and 'folows' for 'follows' in Theorem 5.2.","section":"Throughout"},{"comment":"The statement 'I(31) = {s ≥ 5 : s odd} or I(31) = {s ≥ 9 : s odd} ∪ {5}' is logically two alternatives; it would be clearer to write that I(31) is one of these two sets, since the 7 ∈ I(31) question is explicitly left open.","section":"Example 4.14"}],"recommendation":"major_revision","confidential_remarks":"The mathematical core of the paper appears defensible, and the overclaim is fixable by completing the small cases or weakening the statements. The main non-presentation concerns are the skipped proof of Theorem 4.2 and the unverified Weil-bound hypothesis in Theorem 3.12; both are load-bearing but seem repairable within the manuscript's scope. I would not recommend rejection unless the authors are unable to supply the missing proofs or to adjust the claims about 'all' codes."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The main result is real and worth knowing: the paper determines the covering radius for generalized Zetterberg codes with q0^s ≡ 7 mod 8, the case left open in [31]. Section 3, where the upper bound of 3 is proved, is the strongest part. The reduction to properties of sum systems, the handling of even and odd indices, and the Weil-sum estimates are detailed and mostly convincing. The s=1 result, the explicit threshold s*, and the extension to quasi-perfect twisted half codes are all useful, and the paper is honest about the heavy reliance on [31] where appropriate.\n\nBut the headline claim is too strong. Example 4.15 states plainly that membership of 7 and 9 in I(47) is undetermined, so the covering radii of C_7(47) and C_9(47) are not known. The abstract and concluding remarks claim all generalized Zetterberg codes are determined; that is not established. The stress-test note is right. This is a stated limitation, not a hidden error, but it means the title and abstract oversell.\n\nThere are three other soft spots, in decreasing order of importance. First, Theorem 4.2, the exact characterization that the covering radius is 3 iff some Property NP_i holds, is stated and then skipped with only a pointer to [31]. It is load-bearing, and even if the methods are identical, a paper that relies on it should either prove it or quote it as a theorem with a precise location. Second, the Weil bound for the degree-six polynomial (1−x^2)Δ(x) is asserted without ruling out the possibility that it is a nonsquare times a square, in which case the character sum could be ≈ −q instead of O(√q). That would break the lower bound in Theorem 3.12. I suspect a short computation closes this, but it is not in the paper. Third, the computational checks are described in words but no Magma code or reproducible details are given, which makes the small-field verification hard to trust.\n\nAlso note a minor internal inconsistency: the introduction credits [31] with introducing twisted half generalized Zetterberg codes, but the abstract and Section 5 claim this paper introduces them. That should be fixed for attribution clarity.\n\nBottom line: this deserves a serious referee. The gaps are repairable, but the paper needs revision before acceptance. Weaken the claim to \"all except possibly a finite list,\" or finish the two unresolved cases; provide a proof or precise reference for Theorem 4.2; justify the Weil bound. If those are addressed, this will be a solid contribution to a genuinely hard problem.","headline":"The Section 3 machinery is solid and the main open case is genuinely addressed, but the paper overclaims completion when Example 4.15 leaves two small covering radii undetermined, and the core Theorem 4.2 is not proved here.","tokens_in":22311,"tokens_out":5295,"would_cite":true,"duration_ms":49478,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["94B05"],"pacs":[],"model":"deepseek-v4-flash","headline":"This paper solves the open problem of determining the covering radius of every generalized Zetterberg code in odd characteristic, proving it is 2 or 3 with an explicit criterion.","keywords":["covering radius","generalized Zetterberg codes","twisted half generalized Zetterberg codes","quasi-perfect codes","Weil sums","algebraic curves over finite fields","finite fields","odd characteristic"],"falsifier":"Check a currently undecided finite case, for instance $q_0=31$, $s=7$ or $q_0=47$, $s=7$: determine whether the system $y_i^2=x^2-\\alpha_i$, $i=1,\\ldots,m$, has a solution with $x,y_1,\\ldots,y_m\\in\\mathbb{F}_{q_0^s}^*$, and compare with the covering radius of $\\mathcal{C}_s(q_0)$ computed by exhaustive search. If the radius is not exactly 3 when the system is solvable, the paper's central dichotomy fails. For the large-$s$ claim, recompute the genus of the fiber-product curve directly; any value different from $1+2^{m-1}(m-2)$ invalidates the Hasse–Weil bound that produces $s^*$.","tokens_in":21196,"feed_emoji":"📡","tokens_out":12391,"duration_ms":92300,"temperature":0.7,"pith_summary":"A code's covering radius is the largest distance any received word can be from the code and still be correctable; knowing it tells you how many errors a decoder must handle. This paper closes the last open case for the generalized Zetterberg codes over finite fields of odd characteristic, the case where the field size $q_0^s$ is congruent to $7$ mod $8$. It proves the covering radius is at most $3$, gives a criterion for when it is exactly $3$ rather than $2$, and shows that all sufficiently large odd extension degrees give radius $3$. It also introduces twisted half generalized Zetterberg codes and proves they inherit the same covering radius, yielding quasi-perfect codes when the radius is $2$. Together with the earlier work [31], this settles the covering radius for every generalized Zetterberg code in odd characteristic.","feed_headline":"Every generalized Zetterberg code now has a known covering radius","feed_subtitle":"Last open case, q^s = 7 mod 8, resolved: covering radius is 2 or 3 and twisted half codes are quasi-perfect.","key_machinery":"The engine is a translation of the covering-radius question into solvability of systems of equations over finite fields, and then into rational points on a fiber-product curve. For the upper bound, the code has covering radius at most $3$ if every nonzero field element is a sum of three elements from the subgroup $H_m$ of $\\mathbb{F}_{q^2}^*$ of order $m(q+1)$, where $m=(q_0-1)/2$; this is the family of “Property $P_i$” conditions indexed by the $2^\\ell$-th roots of unity. Each such condition reduces to finding $x_1,x_2,x_3,y_1,y_2,y_3 \\in \\mathbb{F}_q$ with prescribed sums and each $x_j^2+y_j^2$ equal to $1$ (or to a fixed nonsquare $D$). Weil sums bound the number of admissible $x_1$'s, with the exceptional values $\\alpha=\\pm1$ treated separately. For exactness, “Property $\\mathrm{NP}_i$” asserts that some $\\gamma$ with $\\gamma^q = \\theta^i \\gamma$ is not a sum of two elements of $H_m$; this is equivalent to the unsolvability of a system that, for even $i$, becomes the existence of $x,y_1,\\ldots,y_m \\in \\mathbb{F}_q^*$ with $y_j^2 = x^2 - \\alpha_j$ for all nonzero squares $\\alpha_j$ in $\\mathbb{F}_{q_0}$, with an analogous nonsquare version for odd $i$. The fiber-product curve $\\chi: y_j^2 = x^2 - \\alpha_j$, $j=1,\\ldots,m$, has genus $1+2^{m-1}(m-2)$; the Hasse–Weil bound on its rational points yields the threshold $s^*$ beyond which solutions certainly exist, forcing radius $3$.","core_discovery":"The paper establishes that when $q_0^s \\equiv 7 \\pmod 8$, equivalently $q_0 \\equiv 2^\\ell - 1 \\pmod{2^{\\ell+1}}$ with $s$ odd, the generalized Zetterberg code $\\mathcal{C}_s(q_0)$ of length $q_0^s+1$ over $\\mathbb{F}_{q_0}$ has covering radius at most $3$ (Corollary 3.14). It is exactly $3$ precisely when at least one of the properties $\\mathrm{NP}_i$ holds, and exactly $2$ otherwise (Theorem 4.2). The case $s=1$ always gives radius $2$ (Corollary 4.7), while every odd $s$ at or above an explicit threshold $s^*$ gives radius $3$ (Theorem 4.8). The threshold comes from a Hasse–Weil lower bound on the number of $\\mathbb{F}_{q_0^s}$-rational points of a fiber-product curve $y_i^2 = x^2 - \\alpha_i$, where $\\alpha_i$ runs through the nonzero squares of $\\mathbb{F}_{q_0}$. The twisted half generalized Zetterberg codes of length $(q_0^s+1)/2$ have the same covering radius as the full codes, and because their packing radius is $1$, they are quasi-perfect exactly when the covering radius is $2$ (Theorem 5.2).","pith_inferences":["The paper leaves implicit that the unresolved small cases it lists, such as $q_0=31$, $s=7$ or $q_0=47$, $s=7,9$, can be decided not by exhaustive column searches but by checking whether the fiber-product curve has an $\\mathbb{F}_{q_0^s}$-rational point above some $u\\in\\mathbb{F}_q$, i.e. whether $N(s)>0$.","If the quoted genus $1+2^{m-1}(m-2)$ is correct, the same Hasse–Weil counting gives explicit thresholds for every base field in the congruence class, and the examples $q_0=7,23,31,47$ are just the smallest instances of a uniform phenomenon.","The dichotomy between radius 2 and radius 3 suggests a sharper structural fact: the covering radius is decided by a single obstruction, namely whether one auxiliary element $\\gamma$ fails to be a sum of two elements of $H_m$, so computing $\\rho$ is equivalent to solving one system of equations rather than an optimization over all received words."],"forward_implications":["For every finite field of odd characteristic and every $s\\ge1$, the covering radius of $\\mathcal{C}_s(q_0)$ is now an explicit value, 2 or 3, decided by whether the appropriate $\\mathrm{NP}_i$ property holds.","When $s=1$, the twisted half code has parameters $[(q_0+1)/2,(q_0-3)/2,3\\le d\\le4]$ and covering radius 2, so it is quasi-perfect.","For each base field with $q_0\\equiv2^\\ell-1\\pmod{2^{\\ell+1}}$, all odd $s\\ge s^*$ give covering radius 3, with $s^*$ defined explicitly by a single inequality.","If an odd extension degree $s$ gives radius 3, then every odd multiple $st$ also gives radius 3 (Proposition 4.11).","Twisted half generalized Zetterberg codes inherit the full code's covering radius, so the construction supplies quasi-perfect codes whenever the radius is 2."],"supporting_citations":[{"why":"Defines the generalized Zetterberg codes, settles all congruence classes except $q_0^s\\equiv7\\pmod8$, poses the open problem solved here, and supplies the genus formula for the fiber-product curve and the half-code methods this paper extends.","marker":"[31]"},{"why":"Supplies Weil's sum bounds and the quadratic character-sum identity used to lower-bound the number of admissible first coordinates in the system of equations.","marker":"[26]"},{"why":"Supplies the Hasse–Weil inequality and algebraic function field background used to lower-bound the number of rational points on the fiber-product curve.","marker":"[17]"},{"why":"The computer algebra system used for the finite-field searches that settle the small fields $q_0=7,23,31,47$ and the small extension degrees the general bounds do not cover.","marker":"[2]"},{"why":"Provides the sphere-packing bound used to show the twisted half codes have minimum distance at most 4 and packing radius 1, so covering radius 2 makes them quasi-perfect.","marker":"[27]"}],"fun_headline_variants":["Zetterberg covering radius solved for all odd characteristic","Open case closed: Zetterberg codes have radius 2 or 3","Twisted half Zetterberg codes quasi-perfect when radius 2","Hasse-Weil bound pins down Zetterberg covering radius","All generalized Zetterberg codes now have known coverage"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is that the auxiliary curve $y_i^2 = x^2 - \\alpha_i$ for $i=1,\\ldots,m$ has exactly the genus $1+2^{m-1}(m-2)$ quoted from the earlier paper, and the entire threshold argument for large $s$ imports that number without recomputing it.","fun_headline_variants_meta":{"raw":{"variants":["Zetterberg covering radius solved for all odd characteristic","Open case closed: Zetterberg codes have radius 2 or 3","Twisted half Zetterberg codes quasi-perfect when radius 2","Hasse-Weil bound pins down Zetterberg covering radius","All generalized Zetterberg codes now have known coverage"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000204,"raw_usage":{"total_tokens":1458,"prompt_tokens":1079,"completion_tokens":379,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":695,"completion_tokens_details":{"reasoning_tokens":292}},"tokens_in":695,"tokens_out":379,"duration_ms":4181,"temperature":1.0,"reasoning_tokens":292,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-12T15:33:51.269412+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Check a currently undecided finite case, for instance $q_0=31$, $s=7$ or $q_0=47$, $s=7$: determine whether the system $y_i^2=x^2-\\alpha_i$, $i=1,\\ldots,m$, has a solution with $x,y_1,\\ldots,y_m\\in\\mathbb{F}_{q_0^s}^*$, and compare with the covering radius of $\\mathcal{C}_s(q_0)$ computed by exhaustive search. If the radius is not exactly 3 when the system is solvable, the paper's central dichotomy fails. For the large-$s$ claim, recompute the genus of the fiber-product curve directly; any value different from $1+2^{m-1}(m-2)$ invalidates the Hasse–Weil bound that produces $s^*$.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Defines the generalized Zetterberg codes, settles all congruence classes except $q_0^s\\equiv7\\pmod8$, poses the open problem solved here, and supplies the genus formula for the fiber-product curve and the half-code methods this paper extends."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies Weil's sum bounds and the quadratic character-sum identity used to lower-bound the number of admissible first coordinates in the system of equations."},{"cited_title":"Garcia, H","cited_arxiv_id":null,"evidence_quote":"Supplies the Hasse–Weil inequality and algebraic function field background used to lower-bound the number of rational points on the fiber-product curve."},{"cited_title":"Bosma, J","cited_arxiv_id":null,"evidence_quote":"The computer algebra system used for the finite-field searches that settle the small fields $q_0=7,23,31,47$ and the small extension degrees the general bounds do not cover."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Provides the sphere-packing bound used to show the twisted half codes have minimum distance at most 4 and packing radius 1, so covering radius 2 makes them quasi-perfect."}],"review_version":1}