{"id":"32383b82-670e-42ec-9a4b-c75366cd3058","arxiv_id":"2608.11145","paper_version":1,"verdict":"ACCEPT","confidence":"MODERATE","novelty_score":8.0,"correctness_risk":"low","formal_verification":"none","parameter_count":0,"one_line_summary":"For a broad parameter range, the Assouad dimension of Weierstrass and Takagi function graphs is strictly less than 2, with explicit quantitative upper bounds.","lead":"This paper proves that the graphs of Weierstrass functions and many generalized versions have Assouad dimension strictly below 2, contradicting earlier expectations. It provides explicit quantitative bounds for classical Weierstrass and Takagi functions, and shows their graphs are porous.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"No significant objection identified. The PD condition is the key assumption, but it is fully verified for the cosine and sawtooth generators via Proposition 6.1, and the counting and Aikawa arguments are internally consistent.","rationale":"The central claim Theorem 1.1 reduces to verifying the b-adic positive difference condition for the cosine generator. That verification is fully carried out through Proposition 6.1, whose proof by contradiction is sound: the no-PD assumption yields the Lipschitz-type bound (6.2), which forces the oscillation inequality that contradicts (6.1). I checked the non-trivial steps in the main counting argument: the existence of the interval I_φ follows from continuity of F(t) and strict positivity of the excess; the index-pairing in Lemma 3.1 is valid, with q an integer between 1 and b^L−1 and the disjointness of {μ, μ+q} justified by I_φ∩(I_φ+k_φ b^{−n_φ})=∅; the iteration in Lemma 3.2 is quantitatively consistent, including the handling of the residual levels via s≥(N−p−M)/L; Lemma 4.1 correctly applies Lemma 3.2 with (p,N,n)=(p,j,m) and the projected slice estimate is sound; Theorem 4.2 then gives the stated Aikawa bound. The explicit cosine bound in Theorem 6.2 satisfies the required inequalities (3.2), and the quantity is strictly below 2 because θ_φ>0. The Takagi result in Theorem 1.3 is fully supported by Proposition 6.1 and Theorem 4.2; the sketch in Theorem 6.3 is only an explicit-formula bonus and is not load-bearing. The reversed ratio typo in Theorem 5.1 is a display error: the correct inequality b^{n−p}<b R/r follows from R≥b^{−p} and r<b^{−n+1}, and it yields the stated constant 10b^{2−θ}C_0. Thus no significant objection to the central claim is warranted.","tokens_in":21256,"tokens_out":34754,"duration_ms":369824,"concrete_test":"Independently re-derive and numerically check Lemma 3.1 for one non-trivial parameter set, e.g. a=(π+1)/5, b=5, φ(t)=cos(2πt), with I_φ and L as in Theorem 6.2: for a grid of k, n, Q, y, verify the bound #{Q'∈G^n_{k+L}(A,y): Q'⊂Q} ≤ (1−|I_φ|/2)b^L. Also confirm numerically that the chosen L satisfies b^L|I_φ|≥2 and a^L(1/(1−a)+2π/(ab−1))≤η/4. A single violation would expose a hidden flaw in the counting; success would confirm the central estimate.","verdict_should_be":"UNCHANGED","load_bearing_attack":"After checking the proof chain, all load-bearing steps check out. The PD condition is indeed what makes Lemma 3.1 work, but for the actual generators of Theorems 1.1 and 1.3 it is established by the self-contained Proposition 6.1: the strict inequality (6.1) holds for cos(2πt) when ab≥π+1 and for the sawtooth when ab≥3. Lemma 3.1's index-pairing argument is valid, including the count #V≥|I_φ|b^L−1, and Lemma 3.2's iteration uses the maximality of s correctly. Lemma 4.1's dyadic annulus estimate is sound, and Theorem 4.2 correctly derives dim_A≤2−θ_φ from the Aikawa integral criterion. The explicit bounds in Theorem 6.2 check out for b≥5, with the conditions (3.2) satisfied by choice of L. The only textual slips are a reversed ratio in a display in Theorem 5.1 (b^{n−p}<b r/R should read b^{n−p}<b R/r; the final bound still follows from the correct inequality) and the abbreviated proof of Theorem 6.3. Neither affects the qualitative central claim, since Theorem 1.3 follows from Proposition 6.1 and Theorem 4.2, not from the explicit formula in Theorem 6.3.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper proves that for parameters a in (0,1) and b in N with ab >= pi+1, the Assouad dimension of the graph of the classical Weierstrass function W_{a,b} is strictly less than 2. The main tool is a generalized Weierstrass function W^phi_{a,b}; under a b-adic positive difference (PD) condition on the generator phi, the author proves dim_A G(W^phi_{a,b}) <= 2 - theta_phi < 2 (Theorem 4.2), with an explicit theta_phi, via the Aikawa dimension characterization. The cosine generator satisfies the PD condition for ab >= pi+1 and the sawtooth generator for ab >= 3, yielding Theorems 1.1 and 1.3; the latter answers negatively Yu's conjecture for Takagi functions in this range. The paper also gives quantitative bounds, porosity corollaries, and upper bounds for horizontal slices.","tokens_in":21529,"tokens_out":32923,"duration_ms":291266,"significance":"If correct, Theorem 1.1 resolves an open question of Fraser and overturns the prevailing heuristic, supported by typical Holder graphs and Wiener processes, that Weierstrass graphs should have Assouad dimension 2. The proof is notable for introducing a checkable b-adic PD condition and for deriving the bound with no fitted constants: theta_phi, I_phi, and L are constructed explicitly from the PD data. The counting argument in Section 3 and the Aikawa integral estimates in Section 4 are internally consistent, and the verifications for the cosine and sawtooth generators are explicit. The negative answer to Yu's conjecture for Takagi functions in the range ab >= 3 is a further concrete advance, as are the porosity and slice dimension corollaries.","major_comments":[],"minor_comments":[{"comment":"The display 'b^{n-p} < b r/R' is reversed; it should read 'b^{n-p} < b R/r'. The final bound N_r <= 10 b^{2-theta} C_0 (R/r)^{1-theta} follows from the corrected inequality, so this is a local typo rather than a substantive flaw.","section":"Section 5.1, proof of Theorem 5.1"},{"comment":"The line 'Since P_n(J) is compact' is not justified for half-open intervals J in D_n, and the implication dist(y,P_n(J)) <= A a^n implies existence of x in J with |P_n(x)-y| <= A a^n can fail at boundary points. This is easily repaired by working with closures of the intervals or by a limiting argument with a slightly enlarged A and correspondingly adjusted M_A, but the step should be addressed in a revision.","section":"Section 3, Lemma 3.1"},{"comment":"The proof of Theorem 6.3 is only a sketch ('Following similar arguments ... it can be shown'). Since the theorem states an explicit quantitative bound, please include the verification that n_phi=1, k_phi=floor(b/3), t_0=0 satisfy the PD condition, that I_phi=[1/150,1/6] satisfies (3.1), and that the chosen L satisfies (3.2).","section":"Theorem 6.3"},{"comment":"The range b in (a^{-1}(pi+1), infinity) should presumably be b >= a^{-1}(pi+1) to include the equality case ab = pi+1 covered by Theorem 1.1.","section":"Corollary 1.5"},{"comment":"There are several small presentation issues: 'we ahve' in the proof of Theorem 4.2, 'sp < theta theta' in Corollary 7.2, 'Cambridge Trats' in reference [21], and the phrase 'at most or equal to' in Theorem 6.2. The displayed formula in Theorem 6.2 is also difficult to parse; stating the bound as 2 - log(13/12)/(L log b) with L defined separately would improve readability.","section":"Various"}],"recommendation":"minor_revision","confidential_remarks":"I concur with the positive assessment of the proof's soundness. The main result is a significant advance and the PD condition is a useful innovation. The issues I found are local and readily fixable; once the reversed inequality in Section 5.1, the half-open interval boundary point, and the sketched proof of Theorem 6.3 are addressed, the paper should be accepted."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The headline is that this paper proves something the field suspected was false: for the classical Weierstrass function W_{a,b}, the Assouad dimension of the graph is strictly less than 2 whenever ab >= pi+1. The main machinery, a b-adic positive difference condition plus Aikawa interval counting, is genuinely new, and the results for generalized Weierstrass functions and Takagi functions are substantial. I agree with the reader's verdict: accept, with moderate confidence.\n\nWhat the paper does well: the proof is mostly self-contained. Lemma 3.1's pairing argument is valid, Lemma 3.2's iteration uses maximality correctly, and Lemma 4.1 plus Theorem 4.2 derive the Aikawa integral bound cleanly. Proposition 6.1 gives a crisp sufficient condition that checks out for the cosine generator when ab >= pi+1 and for the sawtooth when ab >= 3. There are no fitted constants, no circular dependence on the claimed theorem, and the explicit bounds in Theorem 6.2 are derived rather than asserted. The citation pattern looks appropriate; the author's own prior work is cited where relevant, and the external Aikawa-Assouad equivalence is properly attributed.\n\nThe soft spots are minor and do not threaten the central claim. Theorem 6.3, the explicit Takagi bound, is only sketched: it says the verification is similar and does not show the full check of I_phi or L. I would want that expanded before relying on the explicit formula, though Theorem 1.3 itself follows from Proposition 6.1 plus Theorem 4.2 without it. There is a reversed ratio in a display in Theorem 5.1 (b^{n-p}<b r/R should be b R/r); the final bound still follows, so it is a typo. The half-open b-adic interval boundary conventions are never discussed; in principle a count could be off by an additive constant depending on whether an endpoint lies in I_phi. Such a constant is absorbed into the existing constants, so this is a presentation gap, not a proof gap. The paper also honestly flags that the range 1<ab<pi+1 remains open and that the bound 2-theta_phi is likely not sharp. Those are limitations, not flaws.\n\nWho this is for: fractal geometers working on Assouad dimension, function graphs, porosity, and related analytic applications. It deserves a serious referee. A good editor should send it out, asking the author to expand the Takagi calculation, fix the small typos, and clarify the interval conventions. Modest revision, not a rewrite.","headline":"First strict upper bound for the Assouad dimension of Weierstrass graphs, dim_A < 2 for ab >= pi+1, with a largely sound proof and only minor gaps in the explicit Takagi bound.","tokens_in":22107,"tokens_out":1955,"would_cite":true,"duration_ms":20749,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["28A80","31E05"],"pacs":[],"model":"deepseek-v4-flash","headline":"The paper proves that the Assouad dimension of classical Weierstrass function graphs is strictly less than 2 for all a ∈ (0,1) and integer b with ab ≥ π+1, disproving the prevailing expectation that it should equal 2, and gives…","keywords":["Assouad dimension","Weierstrass functions","Takagi functions","Aikawa dimension","porosity","fractal graphs","positive difference condition","horizontal slices"],"falsifier":"Compute the local covering exponent of G(W_{a,b}) for a specific parameter pair in the range, say a=0.95 and b=5: if as r→0 the number of r-balls needed to cover a ball of radius R on the graph grows like (R/r)^2, then dim_A = 2 and Theorem 1.1 would be false. As a preliminary check, the one-line (PD) inequality for this pair, 2π/(ab−1) < 4/(1+a), can be evaluated directly to confirm the proof's premise holds.","tokens_in":21026,"feed_emoji":"📐","tokens_out":9212,"duration_ms":74726,"temperature":0.7,"pith_summary":"This paper aims to settle the Assouad dimension of Weierstrass function graphs, which had been an open problem with many signs pointing toward the value 2. Under the condition ab ≥ π+1 it proves dim_A G(W_{a,b}) < 2 for the classical Weierstrass function, and more generally proves dim_A G(W^φ_{a,b}) ≤ 2−θ_φ < 2 for any Lipschitz 1-periodic generator φ satisfying a b-adic positive difference condition, with θ_φ explicit. The same result gives dim_A < 2 for Takagi function graphs when ab ≥ 3, contradicting a conjecture stated in [42]. The proof runs through the equivalence between the Aikawa and Assouad dimensions, counting the b-adic intervals at each scale that lie close to the graph and converting the count into an integral bound. If correct, the paper delivers the first non-trivial upper bound for these graphs, along with porosity and analytic corollaries such as Muckenhoupt and Hardy-type inequalities.","feed_headline":"Weierstrass graphs have Assouad dimension strictly below 2","feed_subtitle":"For ab ≥ π+1, a quantitative bound 2−θ overturns earlier hints that the value should be 2.","key_machinery":"The load-bearing mechanism is the b-adic positive difference (PD) condition: some truncation P_{n_φ} has a b-adic secant satisfying |P_{n_φ}(t_0+k_φ $b^{{−n_φ}}$) − P_{n_φ}(t_0)| > κ_φ k_φ $b^{{−n_φ}}$, meaning the secant is steeper than the maximal slope κ_φ that any combination of lower-frequency terms can contribute. Because φ is 1-periodic, the same strict separation persists in every longer truncation on an interval I_φ. The proof then counts the b-adic intervals at scale $b^{{−n}}$ lying close to the graph (the 'Aikawa intervals'), showing at most C $b^{{(1−θ_φ)(N−p)}}$ of them occur inside any level-p interval, and converts this count, through the known equivalence between Aikawa and Assouad dimensions, into an integral estimate on dist(z,G)^{−ϵ} that forces dim_A ≤ 2−θ_φ. Separately, a simple sufficient criterion, Lip(φ)/(ab−1) < 2 osc φ/(1+a), verifies the (PD) condition for the cosine and sawtooth generators.","core_discovery":"On the paper's own terms: for a ∈ (0,1), b ∈ N with ab > 1 and a Lipschitz 1-periodic φ, define W^φ_{a,b} = Σ_{j=0}^∞ a^j φ(b^j x). If some finite truncation P_{n_φ} has a b-adic secant slope strictly larger than κ_φ = Lip(φ)/(ab−1), persisting across an interval I_φ, then dim_A G(W^φ_{a,b}) ≤ 2−θ_φ < 2, where θ_φ = −log(1−|I_φ|/2)/(L log b) for a suitable level L. For φ(t) = cos(2πt) the condition holds when ab ≥ π+1, giving dim_A G(W_{a,b}) < 2; for the sawtooth φ(t) = dist(t,Z) it holds when ab ≥ 3, giving dim_A G(T_{a,b}) < 2. The proof also yields dim_A W_y ≤ 1−θ_φ for every horizontal slice and porosity of the Weierstrass graph, with consequences for Muckenhoupt weights and Hardy inequalities.","pith_inferences":["The (PD) condition is a steep-secant hypothesis; a natural guess is that any generating function whose scaled increments are genuinely one-sided at b-adic scales should satisfy it, which would extend the theorem beyond cosine and sawtooth to, for example, random or generic Lipschitz phases.","Because θ_φ is built from the length of the interval I_φ, optimizing the choices of n_φ and k_φ could yield dimension bounds closer to the true value; the paper does not claim sharpness, so the explicit bounds in Theorems 6.2–6.3 should be read as first quantitative estimates.","If the counting argument can be pushed to weaker differences, such as secants that barely exceed κ_φ on a thin set, then the range 1 < ab < π+1 for the classical Weierstrass function could become accessible, since the obstacle is only the verification of the (PD) condition in that range.","The slice statement dim_A W_y < 1 combined with the porosity of the graph suggests that level sets of these Weierstrass-type functions are strongly homogeneous; one could test computationally whether the bound in Theorem 1.4 is attained, which the paper does not address."],"forward_implications":["For every a ∈ (0,1) and integer b with ab ≥ π+1, the classical Weierstrass graph has Assouad dimension strictly below 2, answering the open question posed in [21] for this parameter range.","Takagi function graphs have Assouad dimension strictly below 2 whenever ab ≥ 3, refuting the conjecture in [42] that the dimension equals 2 in that range.","Every horizontal slice W_y = {x ∈ [0,1] : W^φ_{a,b}(x) = y} satisfies dim_A W_y ≤ 1−θ_φ < 1, so all slices are porous subsets of R.","The Weierstrass graph is porous (Corollary 1.5), and as a result the distance weight dist(z,G)^{−α} belongs to the Muckenhoupt class A_p exactly when (1−p)(2−dim_A G) < α < 2−dim_A G; related Triebel–Lizorkin Hardy inequalities hold for sp < θ_φ.","The bounds are quantitative: for the cosine generator, dim_A G(W_{a,b}) ≤ 2 − log(13/12)/(L log b) with L given explicitly, and for Takagi functions a similar bound uses log(25/23)."],"supporting_citations":[{"why":"Supplies the theorem that Aikawa dimension equals Assouad dimension, the bridge converting counting estimates into a dimension bound.","marker":"[32]"},{"why":"Poses the open question on the Assouad dimension of Weierstrass graphs that Theorem 1.1 answers.","marker":"[21]"},{"why":"Proves for b=2 and a∈(1/2,1) that Takagi graphs have Assouad dimension below 2, the benchmark the paper extends to ab≥3.","marker":"[4]"},{"why":"States the conjecture that Takagi graphs have Assouad dimension 2 and provides the weak-tangent and level-set results the paper refutes in the ab≥3 range.","marker":"[42]"},{"why":"Gives the equivalence between Assouad dimension < d and porosity used to derive Corollary 1.5.","marker":"[33]"},{"why":"Determines the Hausdorff dimension of Weierstrass graphs, the companion result that motivated the search for the Assouad dimension.","marker":"[38]"},{"why":"Supplies the A_p characterization for distance weights used in Corollary 7.1.","marker":"[18]"},{"why":"Supplies the Hardy-inequality equivalence used in Corollary 7.2.","marker":"[27]"}],"fun_headline_variants":["Weierstrass graphs: Assouad dimension strictly below 2","Assouad dimension of Weierstrass graphs is not 2","New quantitative bound: Weierstrass graph Assouad dim < 2","Weierstrass and Takagi graphs: Assouad dim < 2"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"Everything rests on the existence of a b-adic interval where a finite truncation has a genuinely steeper slope than the maximum slope all lower-frequency terms can contribute; if no such interval exists for a given generating function, the counting estimates and the dimension bound do not apply.","fun_headline_variants_meta":{"raw":{"variants":["Weierstrass graphs: Assouad dimension strictly below 2","Assouad dimension of Weierstrass graphs is not 2","New quantitative bound: Weierstrass graph Assouad dim < 2","Weierstrass and Takagi graphs: Assouad dim < 2"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000689,"raw_usage":{"total_tokens":3213,"prompt_tokens":1131,"completion_tokens":2082,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":747,"completion_tokens_details":{"reasoning_tokens":1999}},"tokens_in":747,"tokens_out":2082,"duration_ms":14023,"temperature":1.0,"reasoning_tokens":1999,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-12T05:18:17.639977+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Compute the local covering exponent of G(W_{a,b}) for a specific parameter pair in the range, say a=0.95 and b=5: if as r→0 the number of r-balls needed to cover a ball of radius R on the graph grows like (R/r)^2, then dim_A = 2 and Theorem 1.1 would be false. As a preliminary check, the one-line (PD) inequality for this pair, 2π/(ab−1) < 4/(1+a), can be evaluated directly to confirm the proof's premise holds.","supporting_citations":[{"cited_title":"A note on the dimensions of A ssouad and A ikawa","cited_arxiv_id":null,"evidence_quote":"Supplies the theorem that Aikawa dimension equals Assouad dimension, the bridge converting counting estimates into a dimension bound."},{"cited_title":"Assouad dimension and fractal geometry , vol","cited_arxiv_id":null,"evidence_quote":"Poses the open question on the Assouad dimension of Weierstrass graphs that Theorem 1.1 answers."},{"cited_title":"Slices of the T akagi function","cited_arxiv_id":null,"evidence_quote":"Proves for b=2 and a∈(1/2,1) that Takagi graphs have Assouad dimension below 2, the benchmark the paper extends to ab≥3."},{"cited_title":"Weak tangent and level sets of T akagi functions","cited_arxiv_id":null,"evidence_quote":"States the conjecture that Takagi graphs have Assouad dimension 2 and provides the weak-tangent and level-set results the paper refutes in the ab≥3 range."},{"cited_title":"Assouad dimension: antifractal metrization, porous sets, and homogeneous measures","cited_arxiv_id":null,"evidence_quote":"Gives the equivalence between Assouad dimension < d and porosity used to derive Corollary 1.5."},{"cited_title":"Hausdorff dimension of the graphs of the classical W eierstrass functions","cited_arxiv_id":null,"evidence_quote":"Determines the Hausdorff dimension of Weierstrass graphs, the companion result that motivated the search for the Assouad dimension."},{"cited_title":"a ck, J., Tuominen, H., and V \\","cited_arxiv_id":null,"evidence_quote":"Supplies the A_p characterization for distance weights used in Corollary 7.1."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the Hardy-inequality equivalence used in Corollary 7.2."}],"review_version":1}