{"id":"fa245d63-8c89-410f-a167-af3c2075f862","arxiv_id":"2502.09950","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":8.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"For CLE_κ the mixing rate exponent is 3κ/8 - 1, and this value transfers to FK percolation under the FK-to-CLE scaling limit conjecture.","lead":"The paper computes the mixing rate exponent for conformal loop ensembles, obtaining 3κ/8 - 1, and shows that, assuming a convergence conjecture, the same exponent governs planar Fortuin-Kasteleyn percolation. This settles a question posed by Duminil-Copin and Manolescu and pins down all near-critical exponents of the model once the conjecture is granted.","discovery_kind":"first_principles","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Theorem 1.3 is not fully proved for κ∈(8/3,4]: Section 5 is an explicit sketch and the κ=4 case is deferred to an unstated continuity argument, despite the unconditional headline claim.","rationale":"The reader's verdict is CONDITIONAL, and my concern does not change it; it sharpens one reason for that condition. The FK-to-CLE reduction is honestly conditional on Conjecture 3.1, and Proposition 1.7 plus Theorem 1.8 are presented with substantial detail. The non-simple CLE computation in Section 4 is also detailed, though it inherits an unproved extension of [SXZ24, Lemma 5.3]. The most load-bearing gap for the unconditional claim is the simple-CLE range: Section 5 is labeled a sketch, the key moduli computation is omitted, and κ=4 is handled by a cited continuity result rather than a proof. Because Theorem 1.3 is advertised as unconditional and Theorem 1.1 needs κ=4, completing Section 5 is necessary before the central claim can be accepted as proved. I am not arguing the result is false; the argument is credible and likely repairable, but the current manuscript does not contain the proof.","tokens_in":35873,"tokens_out":4735,"duration_ms":46882,"concrete_test":"Re-derive the omitted step from Proposition 5.2 to Theorem 5.4: apply Proposition 4.19 to the boundary-length transforms (5.4)-(5.5) and invert the Laplace transform to obtain the measures m1,m2, then compare them term-for-term with (4.22)-(4.23). If the inversion produces any expression different from Z_odd/Z_even, or if the r→0 expansion of Z_odd/Z_even deviates from 1+4cos((κ-4)π/4)r^{3κ/8-1}+O(r^{2(3κ/8-1)}), then Theorem 1.3 fails in the simple-CLE range.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The paper's unconditional Theorem 1.3 covers κ∈(8/3,8)\\{6}, but its proof for κ∈(8/3,4] rests on Section 5, which is explicitly a sketch. Proposition 5.2 quotes [ARS22, Lemmas 4.8 and 5.5], Theorem 5.4 simply states the moduli measures with 'the detailed steps of calculation are omitted', and the κ=4 case is dismissed by 'sending κ↑4' with continuity cited from [ACSW24, Appendix A]. Since Theorem 1.3 for κ<4 is a headline unconditional claim and Theorem 1.1 uses κ=4 (q=4), the central result is not yet established on this range. The asymptotic sign and exponent depend on the exact Z_odd/Z_even formula; a missing factor or sign in the omitted inverse-Laplace step would change the mixing-rate exponent. This is a proof completeness gap, not a detected mathematical error.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper defines a conformal-loop-ensemble (CLE) analog of the Fortuin-Kasteleyn mixing rate and proves that, for κ ∈ (8/3,8) \\ {6}, the CLE mixing rate exponent exists and equals 3κ/8−1, with zero mixing rate at κ=6. Assuming convergence of critical FK_q percolation loop configurations to CLE_κ (Conjecture 3.1), the authors derive the FK mixing rate exponent for q ∈ (1,4] and thereby answer Question 3 of Duminil-Copin and Manolescu; for q=2 the result is unconditional. The CLE computation is based on an exact Radon-Nikodym derivative between the laws of odd-level and even-level CLE loops obtained via Liouville quantum gravity conformal welding and Liouville CFT partition functions, together with a lattice argument (Proposition 1.7) relating the FK mixing rate to a primal/dual circuit event.","tokens_in":36088,"tokens_out":9750,"duration_ms":101052,"significance":"If fully substantiated, the main result is a significant and parameter-free derivation: the exponent 3κ/8−1 is extracted from the exact local expansion of the Radon-Nikodym derivative (1.5), not fitted to a target value. The paper also gives a complete-looking lattice proof of Proposition 1.7, and the q=2 corollary is genuinely unconditional. The connection to Cardy's Coulomb-gas annulus partition functions is appealing and likely to be influential. However, as written the unconditional continuum theorem is not proved on the whole claimed range: the simple-loop case κ ∈ (8/3,4] is only sketched, and the κ=4 case is passed over by a cited continuity argument. These gaps affect the headline claims, so the paper currently needs substantial revision before it can be accepted.","major_comments":[{"comment":"Theorem 1.3 is stated as an unconditional result for all κ ∈ (8/3,8) \\ {6}, and Theorem 1.1 uses κ=4 (q=4). The proof for κ ∈ (8/3,4] is not complete in the manuscript: Proposition 5.2 imports the key boundary-length identity (5.6) from [ARS22, Lemmas 4.8 and 5.5], Theorem 5.4 states that 'the detailed steps of calculation are omitted', and the κ=4 case is disposed of by 'sending κ↑4' with continuity cited from [ACSW24, Appendix A]. The omitted inverse-Laplace calculation that converts (5.4)–(5.5) into the modulus densities em1 and em2 is load-bearing: the exponent 3κ/8−1 and its sign come from the exact Zodd/Zeven ratio, and the κ=4 limit is needed for q=4. Please supply the missing calculation and a proof of the κ=4 limiting statement, or explicitly weaken the theorem to the range for which a complete proof is given.","section":"Section 5 and Theorem 1.3"},{"comment":"The proof of the non-simple case κ ∈ (4,8) rests on identity (4.15), quoted from [SXZ24, Lemma 5.3]. The footnote asserts that the lemma, originally stated for γ=√(8/3), is valid for all γ ∈ (√2,2), but no proof or reference for this extension is provided. This extension is used in Proposition 4.14 and then in Propositions 4.15–4.16 and Theorem 4.20 to obtain the exact ratio (4.32)/(1.5). Since no alternative derivation is supplied, this is a load-bearing unproved input; it should be proved in this paper or replaced by a precise citation covering the full range γ ∈ (√2,2).","section":"Footnote to Proposition 4.14"},{"comment":"The passage from Conjecture 3.1 to the inclusions Aeven_{ε^{1−θ},(1+δ)ε^{1+θ}} ⊆ A(εR;δ) ⊆ Aeven_{ε^{1+θ},(1+δ)ε^{1−θ}} is asserted without proof. The loop metric d defined in Section 3.1 gives proximity of individual loops, but it does not by itself control the correspondence of nesting parity for all loops surrounding the origin, nor does the statement of Conjecture 3.1 provide the simultaneous 'all loops ... corresponding CLE loop' coupling used in the argument. Since Theorem 1.8 is the bridge between the CLE computation and the FK mixing exponent, the proof needs either a stronger convergence statement or a separate lemma establishing convergence of the events Aodd and Aeven under the assumed topology; otherwise the comparison is not checkable.","section":"Proof of Theorem 1.8, Section 1.3"}],"minor_comments":[{"comment":"The sentence 'Let h2 be a random generalized function with the same law as h2 defined above' is self-referential as written; the notation for the mean-zero component should be clarified.","section":"Definition 4.2"},{"comment":"After (1.6), the notation Δp,q(r,R) ≍ Δp,q(R)/Δp,q(r) uses Δp,q(R) without explicitly recalling that it is the same as Δp,q(R) in (1.1); please make the notation consistent.","section":"Section 1.3"},{"comment":"The displayed identity after the Poisson summation contains expressions such as 'gπ/τ' and 'τ/gπ' that are easy to misread; inserting parentheses, for example exp(−(πg/τ)p^2 + ((1−g)π/τ + i(2χ+2πm))p), would improve readability.","section":"Equation (4.30)"}],"recommendation":"major_revision","confidential_remarks":"The paper's own contribution—the exact Radon-Nikodym formula and its consequence for the FK mixing exponent—is significant, and I found no evidence of circularity: the exponent is not fitted. My main concern is that the unconditional theorem for κ ∈ (8/3,4] and the κ=4 case are not actually proved in the manuscript, and the non-simple case depends on an unproved extension of [SXZ24, Lemma 5.3]. These are fixable by adding proofs or making the statements conditional, but they are load-bearing for the paper's central claims. I would also ask the editor to ensure that the reliance on several very recent, overlapping preprints is reviewed carefully."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Two things to know. First, the CLE mixing rate exponent 3κ/8−1 and the exact Radon-Nikodym formula (1.5) are real, new results, and the LQG welding derivation is the first rigorous confirmation of the Coulomb gas/CFT exponent. Second, the unconditional Theorem 1.3 is not yet fully proved on the range κ∈(8/3,4]: Section 5 is an admitted sketch with omitted calculations, and κ=4 is deferred to a cited continuity result. The headline claim runs ahead of the proof on that range.\n\nWhat is actually new: the exact formula comparing odd- and even-level CLE loop laws, the mixing rate exponent, and the transfer theorem (Proposition 1.7 plus Theorem 1.8) linking the FK mixing rate to the CLE event A(r;δ). The FK-side reduction is solid—the modified coupling in Proposition 2.7 that handles the non-increasing event A(r;δ) is a genuinely clever piece of work. For FK-Ising, this yields an unconditional new proof of ι=1. The exponent is not fitted: it comes out of the r→0 expansion of an exact formula, so the circularity burden is low. The paper is also honest about its dependence on Conjecture 3.1.\n\nSoft spots, in order of seriousness. The stress-test note lands: Section 5 explicitly says the detailed steps are omitted, and the κ=4 case is dispatched by \"sending κ↑4\" with continuity cited from [ACSW24, Appendix A]. Since Theorem 1.3 is stated unconditionally for this whole range, the paper needs that calculation written out, or the theorem restated with Section 5 flagged as a proof outline. This is a proof-completeness gap, not an error I can see, and it is closable. Second, the use of [SXZ24, Lemma 5.3] beyond its stated γ range is acknowledged in a footnote but not proved; that dependency should be either proved or explicitly conjectured. Third, the paper leans on several very recent preprints by overlapping authors; the cited results look like prior theorems rather than circular inputs, so I would not call this a circularity problem, but the referee should verify those dependencies carefully.\n\nWho this is for: people working on CLE, LQG, and two-dimensional critical exponents. The FK-Ising corollary and the conditional FK result matter for the Duminil-Copin–Manolescu program. It deserves a serious referee: the core computation is credible and important, and the gaps are concrete and fixable. A referee should insist on a complete Section 5 and a proof of the extended [SXZ24] lemma before acceptance.","headline":"Genuinely new CLE computation that answers Duminil-Copin and Manolescu's Question 3; the unconditional κ∈(8/3,4] claim runs ahead of the proof, which is sketched there.","tokens_in":36641,"tokens_out":2080,"would_cite":true,"duration_ms":20533,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["60K35","60J67","82B20"],"pacs":[],"model":"deepseek-v4-flash","headline":"The mixing rate exponent of FK percolation is 3κ/8 − 1, matching CLE.","keywords":["FK percolation","random-cluster model","mixing rate exponent","conformal loop ensemble","Liouville quantum gravity","conformal welding","near-critical exponents","scaling relations"],"falsifier":"Simulate nested CLE_κ for κ = 4.5 on a fine grid and estimate, for a thin annulus [ε, (1 + δ)ε], the probability difference P[Aodd] − P[Aeven] renormalized by P[Aeven] across ε → 0; if the log-log slope fails to approach 3(4.5)/8 − 1 = 0.6875, Theorem 1.3 is false. For the lattice statement, repeat the analogous estimate on critical FK at q = 3 (κ = 4.8) in Λ_R and check whether log ∆(R) / log R tends to 0.8; a different limit would contradict Theorem 1.1 if the convergence conjecture holds.","tokens_in":35651,"feed_emoji":"🌀","tokens_out":9742,"duration_ms":92837,"temperature":0.7,"pith_summary":"This paper answers Question 3 of the scaling-relations program for planar FK percolation by computing the mixing rate exponent—the power-law rate at which a boundary condition's influence on a far-away edge decays—and proving it equals 3κ/8 − 1, where κ is the CLE parameter attached to the cluster weight q. The lattice computation is conditional on the standard but unproved conjecture that critical FK interfaces converge to nested CLE_κ; for the FK-Ising case q = 2, where convergence is proved, the result is unconditional and gives exponent 1. The paper's genuinely unconditional result lives in the continuum: for CLE_κ with κ ∈ (8/3, 8) \\ {6}, the CLE mixing rate exponent exists and equals 3κ/8 − 1, with the odd-even discrepancy positive below κ = 6, zero at κ = 6, and negative above. If the conjectural convergence holds, plugging this exponent into the earlier scaling relations yields the near-critical exponents of FK percolation.","feed_headline":"CLE loop mixing exponent proven: 3κ/8 − 1","feed_subtitle":"The FK percolation analogue follows under the standard convergence conjecture, giving near-critical exponents.","key_machinery":"The central object is the conformal loop ensemble (CLE_κ), a random collection of noncrossing loops parameterized by κ ∈ (8/3, 8); the mixing rate is the renormalized difference in probability that a thin annulus contains an odd-level versus even-level surrounding loop. The key identity is Theorem 1.5: the Radon-Nikodym derivative of the odd-level loop law against the even-level law equals a ratio of alternating sine-weighted partition functions in the annulus modulus r, with leading term 4 cos((κ − 4)π/4) $r^{{3κ/8 − 1}}$, which determines the exponent. This formula is derived by welding CLE loops to Liouville quantum gravity surfaces: each loop is realized as the conformal welding interface between a generalized quantum disk and a generalized quantum annulus, and the boundary-length laws of these annuli are computed using exact solvability of Liouville field theory. The FK-to-CLE comparison is carried by an annulus event of a primal circuit surrounding a dual circuit, which converges to the CLE odd/even loop events under the conjectured convergence.","core_discovery":"Theorem 1.3 is the unconditional anchor: for each κ ∈ (8/3, 8) \\ {6}, the nested CLE_κ mixing rate has the power-law behavior Δ_κ(r; δ) ≍ $r^{{3κ/8 − 1}}$, with the sign of Δ_κ matching the sign of 6 − κ, and Δ_κ = 0 at κ = 6. Theorem 1.1 then transfers this to FK percolation: for q ∈ (1, 4] with κ = 4π / arccos(−√q / 2) ∈ [4, 6), assuming the convergence conjecture, the lattice mixing rate exponent exists and equals 3κ/8 − 1. The bridge is Proposition 1.7, which shows that a CLE-friendly annulus event—a primal circuit enclosing a dual circuit—captures the same mixing rate as the original edge and crossing events, and Theorem 1.8, which uses the conjectured convergence to compare the two rates. The CLE result comes from an exact formula for the Radon-Nikodym derivative between the laws of odd- and even-nesting CLE loops, expressed as a ratio of $\\theta$-like partition functions.","pith_inferences":["If the same LQG welding machinery were pushed to boundary-touching loops and to loops at arbitrary nesting levels, the exact Radon-Nikodym formula would likely yield explicit annulus crossing probabilities and multi-loop correlation functions; the paper only sketches that the approach extends.","A numerical extraction of the log-log slope of the lattice mixing rate for a value of q without a convergence proof, e.g., q = 3, compared with 3κ/8 − 1 would give an indirect test of the convergence conjecture.","The sign flip at κ = 6 means that odd nesting levels dominate in thin annuli when κ < 6 and even levels dominate when κ > 6; this parity asymmetry vanishes exactly at Bernoulli percolation, matching the known zero mixing rate there."],"forward_implications":["For q ∈ (1, 4], assuming the convergence conjecture, the FK mixing rate exponent is ι(q) = 3κ/8 − 1, settling Question 3 of [DCM22].","Combined with the scaling relations of [DCM22], the near-critical exponents β, γ, and δ are determined for q ∈ (1, 4]; α is also determined for q ∈ [2, 4], and for q ∈ (1, 2) if a further scaling relation holds.","For the FK-Ising model (q = 2), the mixing rate exponent is unconditionally 1.","For CLE_κ with κ ∈ (8/3, 8) \\ {6}, the mixing rate exponent exists and equals 3κ/8 − 1, and the odd-even probability difference changes sign at κ = 6.","The exact comparison of odd/even loop laws reproduces the Coulomb-gas annulus partition functions of the O(n) loop model."],"supporting_citations":[{"why":"Defines the mixing rate exponent, proves the scaling relations it feeds into, and poses the question answered here.","marker":"[DCM22]"},{"why":"Introduces quantum disks and forested lines, the surface types used in the CLE-LQG welding description.","marker":"[DMS21]"},{"why":"Supplies the conformal-radius increment process used to show that multi-loop contributions are negligible in a thin annulus.","marker":"[SSW09]"},{"why":"Provides the extraction of annulus moduli from Liouville field boundary-length distributions, a key input for the modulus laws.","marker":"[ARS22]"},{"why":"Supplies the pinched quantum annulus and forested-line boundary-length formulas used to compute the odd/even loop partition functions.","marker":"[SXZ24]"},{"why":"Gives the symmetry and continuity results for the quantum annulus measures needed in the welding and in the κ → 4 limit.","marker":"[ACSW24]"},{"why":"Gives the Coulomb-gas annulus partition functions whose form the exact odd/even formula reproduces.","marker":"[Car06]"},{"why":"Proves conformal invariance of FK-Ising interfaces, making the q = 2 mixing-rate result unconditional.","marker":"[Smi10]"},{"why":"Completes the FK-Ising scaling-limit convergence to CLE_{16/3}, used in the unconditional corollary.","marker":"[KS19]"}],"fun_headline_variants":["Exact CLE mixing exponent: 3κ/8 − 1 for all κ in (8/3,8)\\{6}","Answer to Question 3: CLE mixing rate equals 3κ/8 − 1","CLE loop mixing law proven—exponent 3κ/8 − 1","From CLE to FK: mixing exponent 3κ/8 − 1 under convergence"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The transfer from the unconditional CLE result to FK percolation rests on the conjecture that critical FK_q interfaces converge to nested CLE_κ, which has been proved only in the FK-Ising case q = 2.","fun_headline_variants_meta":{"raw":{"variants":["Exact CLE mixing exponent: 3κ/8 − 1 for all κ in (8/3,8)\\{6}","Answer to Question 3: CLE mixing rate equals 3κ/8 − 1","CLE loop mixing law proven—exponent 3κ/8 − 1","From CLE to FK: mixing exponent 3κ/8 − 1 under convergence"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000412,"raw_usage":{"total_tokens":2157,"prompt_tokens":992,"completion_tokens":1165,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":608,"completion_tokens_details":{"reasoning_tokens":1062}},"tokens_in":608,"tokens_out":1165,"duration_ms":10472,"temperature":1.0,"reasoning_tokens":1062,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-07T19:58:34.246176+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Simulate nested CLE_κ for κ = 4.5 on a fine grid and estimate, for a thin annulus [ε, (1 + δ)ε], the probability difference P[Aodd] − P[Aeven] renormalized by P[Aeven] across ε → 0; if the log-log slope fails to approach 3(4.5)/8 − 1 = 0.6875, Theorem 1.3 is false. For the lattice statement, repeat the analogous estimate on critical FK at q = 3 (κ = 4.8) in Λ_R and check whether log ∆(R) / log R tends to 0.8; a different limit would contradict Theorem 1.1 if the convergence conjecture holds.","supporting_citations":[],"review_version":1}