{"id":"f424c83d-b2ee-401f-9b8e-d8dbbf755f71","arxiv_id":"2411.15750","paper_version":1,"verdict":"REJECT","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"high","formal_verification":"none","parameter_count":0,"one_line_summary":"A new rational-trace construction is claimed to yield three Kloosterman-sum-characterized classes of bent functions, but the stated theorems omit a necessary trace condition and are false as written.","lead":"This paper introduces rational trace functions as building blocks for Dillon-like bent functions and claims to characterize three classes via Kloosterman sums. As stated, two of the three characterization theorems are false because they omit a trace-zero condition that the proofs silently assume.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"No significant objection identified: the reader's claimed counterexample to Theorem III.2(1) is invalid because Tr^{2n}_1(a1)=0 for a1∈F_q, so h2(0)=0 holds regardless of a2.","rationale":"The paper's central claim is that the rational-trace building blocks in H yield three characterized classes of bent functions, with bentness equivalent to conditions involving Kloosterman sums. The reader's rejection rests on a single alleged error: that the proof of Theorem III.2(1) assumes Tr^{2n}_1(a2)=0 for a2 outside F_q. That attack does not land. In the theorem, a1 is in F_q^*, so Tr^{2n}_1(a1)=a1+a1^q=0, making the asserted product h2(0)=0 true independently of a2. The rest of the b=1 argument in Theorem III.2 is consistent with Eq. (II.7): for a2∉F_q, ξ(a2,1)=1 and ξ(a1+a2,1)=1, so Eq. (III.3) reduces exactly to the stated condition Tr^n_1(a1)=1. I also checked the structure of Theorems III.2(2)-(3) and III.3: the case splits by whether relative traces vanish, the Θ computations match the stated trace conditions, and the b∉U bounds are legitimate for n≥6 with the stated Kloosterman-sum bound. The only presentation-level caveat is the pole convention for x^{q-1}+b=0; this is explicitly flagged in Lemma II.4 and is standard in such exponential-sum computations. Since the central argument appears to hold and the specific counterexample in the reader's verdict is based on a false premise, no significant objection is identified.","tokens_in":22408,"tokens_out":35902,"duration_ms":299497,"concrete_test":"Directly verify the alleged counterexample for Theorem III.2(1) at minimal size: take n=3, q=8, choose a1∈F_8 with Tr^3_1(a1)=1, and choose a2∈F_64\\F_8 with Tr^6_1(a2)=1, the case the reader worries about. Enumerate U={λ∈F_64: λ^9=1}, compute T=Σ_{λ∈U}(-1)^{Tr^6_1(a1/(λ+1))·Tr^6_1(a2/(λ+1))} with the convention 1/0=0, and check T=1=(-1)^{h2(0)}. If T=1, Proposition II.4 confirms h2 is bent, so the claimed failure does not occur.","verdict_should_be":"UNCHANGED","load_bearing_attack":"After checking the central argument, I find no load-bearing error in The stated characterizations. The reader's weakest assumption is not correct. In Theorem III.2(1), the proof uses h2(0)=Tr^{2n}_1(a1)Tr^{2n}_1(a2)=0. Since the theorem assumes a1∈F_q^*, the absolute trace is Tr^{2n}_1(a1)=Tr^n_1(a1+a1^q)=Tr^n_1(0)=0. Hence h2(0)=0 is true for every a2∈F_{q^2}, including a2∉F_q with Tr^{2n}_1(a2)=1. The condition a2∉F_q enters only through the ξ-values: with b=1, Eq. (II.7) gives ξ(a2,1)=ξ(a1+a2,1)=1, and Eq. (III.3) yields Θ=2+q(1+(-1)^{Tr^n_1(a1)})=2 when Tr^n_1(a1)=1, so T=(-1)^{h2(0)}=1 and bentness follows by Proposition II.4. The analogous h3(0) computations in Theorem III.3 also rely on a1∈F_q and are correct. The only genuine caveat is that the functions have poles at x^{q-1}=b, so the paper must use the convention 1/0=0 stated in Lemma II.4; with that convention the proofs are coherent.","agreement_with_reader":"disagree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper introduces a family H of rational trace functions on F_{2^{2n}} of the form Tr^{2n}_1(a/(x^{2^n-1}+b)) and studies the bentness of Dillon-like Boolean functions built from these blocks via reduced polynomials F. It states an explicit exponential-sum formula (Lemma II.5) for ξ(a,b) and uses it to characterize three classes: h1=Tr^{2n}_1(a1/(x^{q-1}+b)) (Theorem III.1), h2=f1 f2 (Theorem III.2), and h3=f1 f2 + f1 f3 + f2 f3 (Theorem III.3). The h1 characterization and the b∉U cases are not affected by the issue discussed below, but the b∈U branches of Theorems III.2 and III.3 rest on an incorrect identity in Lemma II.5. The specific reviewer objection that h2(0) need not vanish in Theorem III.2(1) does not land: for a1∈F_q one has Tr^{2n}_1(a1)=Tr^n_1(a1+a1^q)=0, so h2(0)=0 automatically. The load-bearing problem lies instead in the claimed equivalence used to prove Eq. (II.7).","tokens_in":22740,"tokens_out":36059,"duration_ms":289255,"significance":"If the characterizations were correct, the paper would be significant: it would give the first bent functions assembled from rational trace blocks, with Kloosterman-sum characterizations and concrete examples of functions apparently not EA-equivalent to known monomial bent functions. The paper does not rely on circular arguments or fitted parameters, and the h1 result and the b∉U parts of the proofs are plausible. However, the central b∈U branch of the key exponential-sum lemma is false, and Theorem III.2 is false as stated for b∈U\\{1}; Theorem III.3 is at least unproved and very likely false in the same branch. Since these are the main claimed characterizations, the paper cannot be accepted in its present form.","major_comments":[{"comment":"The proof of Lemma II.5 uses the assertion that for b∈U, a^{q-1}b^2=1 is equivalent to Tr^{2n}_n(ab)=0. This is false for b≠1. For a concrete counterexample with q=4, let ζ be a primitive 5th root of unity in F_{16}, set b=ζ and a=ζ^{-1}. Then ab=1, so Tr^{4}_2(ab)=0; but a^{q-1}b^2 = ζ^{-3}ζ^2 = ζ^4 ≠ 1. Consequently Eq. (II.7) as printed is wrong: in this example the first line would give ξ(a,b)=1+(-1)^{Tr^2_1(a\\bar a)}q = 1+4 = 5, while the reduction ξ(a,b)=ξ(a\\bar a, a^{q-1}b^2) with a\\bar a=1 and a^{q-1}b^2∈U\\{1} gives ξ=1 by Eq. (II.8). The correct case distinction for b∈U depends on c=a^{q-1}b^2 (c=1 versus c≠1), not on Tr^{2n}_n(ab); for b∈U\\{1} and Tr^{2n}_n(ab)=0 one always has c≠1, hence ξ(a,b)=1.","section":"Section II, Lemma II.5, Eq. (II.7)"},{"comment":"Because of the error in Eq. (II.7), the proof of Theorem III.2 in the case b∈U\\{1} is invalid, and conditions (2) and (3) are false. With the correct ξ-values, every a satisfying Tr^{2n}_n(ab)=0 (when b∈U\\{1}) has ξ(a,b)=1. In the situation of conditions (2) and (3) this makes ξ(a1,b)=ξ(a2,b)=ξ(a1+a2,b)=1, so the left-hand side of Eq. (III.3) equals q+1+1+1−1 = q+2, which cannot equal 2(−1)^{h2(0)} = ±2 for q≥2. Thus h2 is not bent under these parameter conditions, contradicting the stated 'if and only if' and the corresponding triples in Example 2.","section":"Section III, Theorem III.2"},{"comment":"The same incorrect ξ-values are used in the b∈U\\{1} subcases of Theorem III.3. In particular, condition (5) requires only Tr^{2n}_n(a2/b)≠0, Tr^{2n}_n(a3/b)≠0 and Tr^{2n}_n((a1+a2+a3)/b)≠0; it does not prevent a2^{q-1}b^2, a3^{q-1}b^2, or (a1+a2+a3)^{q-1}b^2 from being 1. When such an element has c=1, its ξ-value is 1+(−1)^{Tr^n_1(a\\bar a)}q rather than 1, so the right-hand side of Eq. (III.6) is not forced to be 2. The proof's claims that the listed conditions are necessary and sufficient for these subcases (also repeated in Remark 1) are therefore not established, and Theorem III.3 as stated is not supported.","section":"Section III, Theorem III.3"}],"minor_comments":[{"comment":"The notation rendered as 'a/a' in the first branch of Eq. (II.7) is ambiguous; it should be a\\bar a (the field-theoretic norm to F_q), and this should be defined explicitly at first use.","section":"Section II, Eq. (II.7)"},{"comment":"Remark 1 states that the conditions in Theorem III.3 are 'sufficient and necessary', but the theorem itself is written only as an 'if' statement; the paper should either strengthen the theorem or weaken the remark.","section":"Section III, Remark 1"},{"comment":"The text says 'Section IV discusses the EA equivalence, and Section IV concludes the paper'; the second reference should be to Section V.","section":"Section IV"}],"recommendation":"reject","confidential_remarks":"The reader's specific counterexample about h2(0) is not valid, but the manuscript has a genuine, independent error in Lemma II.5 that invalidates the b∈U branches of the two main product theorems. Since the characterization claims are the central contribution and would require reworking the exponential-sum formula and re-deriving the theorems, I recommend rejection rather than minor revision."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Bottom line: the reader's rejection is built on a false counterexample. In the proof of Theorem III.2(1), the assertion h2(0)=0 does not require Tr^{2n}_1(a2)=0. It requires only Tr^{2n}_1(a1)=0, which holds because a1 ∈ F_q^*: Tr^{2n}_1(a1)=Tr^n_1(a1+a1^q)=0. So the product is zero for every a2, including a2 outside F_q with absolute trace 1. Theorem III.3(1) is the same story: both a1 and a2 lie in F_q, so their absolute traces vanish, and h3(0)=0 regardless of a3. The stated conditions are consistent with the proofs.\n\nWhat is genuinely new: the family H of rational trace functions as building blocks; the reduction of bentness of {f_i} combinations to Kloosterman sums via Lemma II.5; and the explicit characterizations in Theorems III.1–III.3, which generalize Li et al.'s criterion. This is, as far as I know, the first time bent functions from rational trace blocks have been characterized in this way. The proofs are long but structured: the heavyweight Lemma II.5 is the crux, and the case analyses in Section III are careful. The paper also gives the ordinary polynomial form and is honest about the limits of its EA-equivalence experiments.\n\nSoft spots: Theorem III.3(6) is a relation between four Kloosterman sums rather than a clean condition, and the paper admits this in Remark 1. The EA-inequivalence evidence is experimental on F_2^6 and F_2^8 only, so the 'new functions' claim is suggestive, not a general proof. The notation is dense; readers outside the finite-field bent-function bubble will need patience.\n\nOverall, I see no fatal flaw. The central argument holds up as far as I've checked, and the reader's strongest objection is a misreading. This is a solid contribution to the bent-function subfield: new families, explicit characterizations, and detailed proofs. I would send it to an expert referee—the exponential-sum calculations deserve close scrutiny, but this is not a desk reject.","headline":"The reader's rejection rests on a misreading—the trace-zero step works because a1 lies in F_q, so the central theorems likely survive and the paper deserves refereeing.","tokens_in":23258,"tokens_out":5980,"would_cite":true,"duration_ms":46290,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["94A60","11T71","06E30","11T23"],"pacs":[],"model":"deepseek-v4-flash","headline":"The paper introduces trace rational blocks and claims explicit Kloosterman-sum characterizations for three classes of Dillon-like bent functions.","keywords":["bent functions","Dillon-like exponents","Kloosterman sums","trace rational functions","Boolean functions","Walsh transform","EA equivalence"],"falsifier":"Take n=3, q=8, b=1, a1 in the subfield with trace 1, and a2 outside the subfield with absolute trace 1; substituting into Eq. (III.3) gives left-hand side 2 and right-hand side -2, so h2 would not be bent even though it satisfies the conditions of Theorem III.2(1).","tokens_in":1978,"feed_emoji":"🧮","tokens_out":4566,"duration_ms":139348,"temperature":0.7,"pith_summary":"Bent functions are Boolean functions on even-dimensional spaces that sit as far as possible from every affine function; they are central to coding theory and cryptography. This paper introduces a family of trace rational blocks $f(x)=\\mathrm{Tr}^{2n}_1(a/(x^{2^n-1}+b))$ on $\\mathbb{F}_{2^{2n}}$ and claims that certain products and sums of these blocks are bent exactly when the parameters satisfy explicit conditions expressed through binary Kloosterman sums. The three classes it characterizes are the single block, a product of two blocks, and the symmetric pairwise sum of three blocks. If these characterizations are correct, they are the first bent functions built from these rational blocks, and they include functions not equivalent to the known monomial bent functions.","feed_headline":"Kloosterman sums characterize three new bent function classes","feed_subtitle":"The paper builds bent functions from rational trace blocks, with criteria written directly in Kloosterman sums.","key_machinery":"The carrying object is the exponential sum $\\xi(a,b)=\\sum_{\\lambda\\in\\mathbb{U}}(-1)^{\\mathrm{Tr}^{2n}_1(a/(\\lambda+b))}$ on the unit circle $\\mathbb{U}=\\{\\lambda\\in\\mathbb{F}_{q^2}:\\lambda^{q+1}=1\\}$. Lemma II.5 evaluates this sum in three regimes — $b\\in\\mathbb{U}$ with $\\mathrm{Tr}^{2n}_n(ab)=0$, $b\\in\\mathbb{U}$ with nonzero trace, and $b\\notin\\mathbb{U}$ — in terms of the binary Kloosterman sum $K_n(a)=\\sum_{x\\in\\mathbb{F}_{2^n}}(-1)^{\\mathrm{Tr}^n_1(1/x+ax)}$. Together with the Walsh-transform reduction of Proposition II.4 (a Dillon-like function is bent iff the signed unit-circle sum equals $(-1)^{f(0)}$), this turns bentness of the block combinations into exact equalities involving Kloosterman sums.","core_discovery":"The central claim is that, for $q=2^n$ and $b\\in\\mathbb{F}_{q^2}^*$, the function $h_2(x)=\\mathrm{Tr}^{2n}_1(a_1/(x^{q-1}+b))\\cdot\\mathrm{Tr}^{2n}_1(a_2/(x^{q-1}+b))$ is bent exactly when one of three parameter conditions holds — for instance $b=1$, $\\mathrm{Tr}^n_1(a_1)=1$, and $a_2\\in\\mathbb{F}_{q^2}\\setminus\\mathbb{F}_q$ — and that the symmetric sum $h_3(x)=\\sum_{i<j}\\mathrm{Tr}^{2n}_1(a_i/(x^{q-1}+b))\\mathrm{Tr}^{2n}_1(a_j/(x^{q-1}+b))$ is bent under six listed conditions. The route is to reduce the Walsh transform of any Dillon-like function to a signed sum over the unit circle $\\mathbb{U}=\\{x:x^{q+1}=1\\}$ (Proposition II.4), then to evaluate the exponential sums $\\xi(a,b)$ that arise, with Lemma II.5 connecting them to the binary Kloosterman sum $K_n$. The paper also derives the ordinary polynomial form of the single block $h_1$, showing it is a full Dillon-type sum, and reports computational checks that the new classes are not EA-equivalent to the five known monomial bent classes.","pith_inferences":["The identity (III.1) is a general Fourier-inversion criterion: for any reduced polynomial $F$, bentness of $F(f_1,\\dots,f_t)$ is equivalent to an explicit linear equation in the $\\xi$-values, so the same machinery can produce characterizations for $F$ of higher degree.","The explicit polynomial form (IV.2) gives a concrete way to test EA-inequivalence beyond the computationally feasible fields: comparing its coefficient set with Eq. (IV.1) on $\\mathbb{F}_{2^{12}}$ or larger would resolve the open question the paper leaves.","Because the range of $K_n(a)$ is completely known, the criteria are finitely checkable; an exhaustive small-$n$ enumeration of the parameter triples and quadruples would let one compile a catalogue of the new bent functions and check their duals, which Proposition II.4 makes easy to compute."],"forward_implications":["If Theorem III.1 holds, the single trace-rational block is bent exactly under the two listed conditions, making $K_n$ the deciding quantity for the $b\\notin\\mathbb{U}$ case.","If Theorem III.2 holds, products of two distinct blocks are bent in exactly the three listed regimes, with the outside-unit-circle regime ruled out for $n\\ge 6$.","If Theorem III.3 holds, the symmetric sum of three blocks admits the six stated parameter families, including a four-Kloosterman-sum relation when $b\\notin\\mathbb{U}$.","The paper's polynomial-form derivation implies the single-block class has every Dillon exponent present, distinguishing it from earlier fixed-term Dillon-like constructions.","The reported equivalence checks indicate the three classes are not EA-equivalent to the known monomial bent classes on the tested fields."],"supporting_citations":[{"why":"gives the Dillon-like bentness criterion that this paper generalizes to the binary rational-trace setting.","marker":"[21]"},{"why":"supplies the Kloosterman-sum range used to bound the non-unit-circle case in Theorem III.2.","marker":"[20]"},{"why":"introduces the hyper-bent/Kloosterman/Dickson-polynomial toolkit that Lemma II.5 builds on.","marker":"[4]"},{"why":"defines the Dillon-like partial-spread class and the Dillon exponent form shared by all functions in the paper's block family.","marker":"[10]"},{"why":"translates Charpin-Gong hyper-bentness into the Boolean-function view used in Proposition II.1.","marker":"[8]"},{"why":"provides the parametrization of the unit circle minus one by the subfield, used throughout the evaluation of the exponential sums.","marker":"[26]"},{"why":"establishes the hyper-bent characterization via the weight vector on the unit circle, the origin of Proposition II.1.","marker":"[30]"}],"fun_headline_variants":["Kloosterman sums pin down three novel bent classes","Rational traces and Kloosterman sums: new bent families","Bent from rational blocks: Kloosterman criterion for new classes","Three classes of bent functions via Kloosterman sums","New bent functions from Dillon-like exponents and Kloosterman sums"],"cache_read_input_tokens":25344,"weakest_assumption_plain":"The b=1 cases of Theorems III.2 and III.3 rely on the premise that any element outside the subfield has absolute trace zero, which forces the initial values h2(0) and h3(0) used in the proof.","fun_headline_variants_meta":{"raw":{"variants":["Kloosterman sums pin down three novel bent classes","Rational traces and Kloosterman sums: new bent families","Bent from rational blocks: Kloosterman criterion for new classes","Three classes of bent functions via Kloosterman sums","New bent functions from Dillon-like exponents and Kloosterman sums"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000209,"raw_usage":{"total_tokens":1455,"prompt_tokens":1040,"completion_tokens":415,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":656,"completion_tokens_details":{"reasoning_tokens":332}},"tokens_in":656,"tokens_out":415,"duration_ms":4319,"temperature":1.0,"reasoning_tokens":332,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-12T13:59:14.980237+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Take n=3, q=8, b=1, a1 in the subfield with trace 1, and a2 outside the subfield with absolute trace 1; substituting into Eq. (III.3) gives left-hand side 2 and right-hand side -2, so h2 would not be bent even though it satisfies the conditions of Theorem III.2(1).","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"gives the Dillon-like bentness criterion that this paper generalizes to the binary rational-trace setting."},{"cited_title":"Lachaud, J","cited_arxiv_id":null,"evidence_quote":"supplies the Kloosterman-sum range used to bound the non-unit-circle case in Theorem III.2."},{"cited_title":"Charpin, G","cited_arxiv_id":null,"evidence_quote":"introduces the hyper-bent/Kloosterman/Dickson-polynomial toolkit that Lemma II.5 builds on."},{"cited_title":"Dillon, Elementary Hadamard difference sets, PhD di ssertation, University of Maryland","cited_arxiv_id":null,"evidence_quote":"defines the Dillon-like partial-spread class and the Dillon exponent form shared by all functions in the paper's block family."},{"cited_title":"Carlet, P","cited_arxiv_id":null,"evidence_quote":"translates Charpin-Gong hyper-bentness into the Boolean-function view used in Proposition II.1."},{"cited_title":"Rosendahl, Niho type cross-correlation functions a nd related equations, PhD dissertation, Univ","cited_arxiv_id":null,"evidence_quote":"provides the parametrization of the unit circle minus one by the subfield, used throughout the evaluation of the exponential sums."},{"cited_title":"Y oussef, G","cited_arxiv_id":null,"evidence_quote":"establishes the hyper-bent characterization via the weight vector on the unit circle, the origin of Proposition II.1."}],"review_version":1}