{"id":"5fa11420-b19a-4f8b-a4b5-3d3eda281436","arxiv_id":"2607.18194","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":8.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":2,"one_line_summary":"For beta1 = o(beta2) both tending to infinity, coupled CDRP free energies at beta1 and beta2 converge to two independent directed landscapes; as a byproduct, the directed landscape is a two-dimensional black noise.","lead":"The paper proves that CDRP free energies at two far-apart inverse temperatures decouple, converging jointly to two independent directed landscapes — the first rigorous temperature-chaos result in the KPZ class — and that the directed landscape is a two-dimensional black noise. A reader should care because temperature chaos is a decades-old physics prediction and the black-noise property resolves a conjecture by Virág.","discovery_kind":"first_principles","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Theorem 1.7's proof applies Lemma 5.4 using only convergence in probability of the restricted free energy, but the lemma requires an L^2 bound that is not established.","rationale":"The reader's weakest assumption was the imported localization estimate [DZ24, Corollary 3.5] (Proposition 4.14), and that is indeed a key external input. My concern is distinct but downstream: even granting Proposition 4.14, the proof of Theorem 1.7 uses it only to prove convergence in probability of Hbar_{β1,S} − Hbar_{β1}, whereas the independence lemma Lemma 5.4 needs L² convergence. This is an internal gap in the visible proof, not an objection to the external result. It is load-bearing because it is the step that turns 'each marginal converges to the directed landscape' into 'the joint limit is a product of two independent directed landscapes.' If the L² bound fails, the central claim of temperature chaos is not established by the given argument. I believe the bound can be recovered by a routine integration of the stretched-exponential tail in Proposition 4.14, and the reader's CONDITIONAL verdict already accommodates the need for such a patch. I therefore do not recommend changing the verdict, but the stated condition should explicitly include this L² strengthening of Lemma 5.1 or an alternative argument in the proof of Theorem 1.7.","tokens_in":79011,"tokens_out":7520,"duration_ms":74995,"concrete_test":"Derive the missing L² bound from Proposition 4.14. Let D_{β1,β2} := Hbar^ξ_{β1,S}(u) − Hbar^ξ_{β1}(u), which equals −(2^{1/3}/β1^{4/3}) log Pξβ1( sup |X| ≤ εβ2^{2/3}). With m = ε(β2/β1)^{2/3} → ∞, (4.11) gives P( Pξ(sup|X| > mβ1^{2/3}) ≥ C1 e^{−m²/C1} ) ≤ C2 e^{−m³/C2}. Use this to bound P(D_{β1,β2} > t) and check whether ∫ t P(D > t) dt → 0 as β1→∞. If this tail integration is written out and yields E|D|² → 0, Theorem 1.7 goes through as stated; if not, Lemma 5.4 is inapplicable and the independence part of Theorem 1.7 is unproved.","verdict_should_be":"UNCHANGED","load_bearing_attack":"In the proof of Theorem 1.7 (Section 5), the authors set X_n = Hbar^ξ_{β1}(u), Y_n = Hbar^ξ_{β2}(u), Z_n = Hbar^ξ_{β1,S}(u), and W_n = Hbar^{η_S}_{β2}(u), and then invoke Lemma 5.4. Lemma 5.4 requires limsup E||X_n − Z_n||² ≤ ε. But the only approximation proved for Z_n is Lemma 5.1, which states that Hbar^ξ_{β1,S}(u) − Hbar^ξ_{β1}(u) converges to 0 in distribution, equivalently in probability. No L² bound is proved for this difference.  The conclusion that the limiting pair (L1,L2) is independent therefore does not formally follow from the stated lemmas, because a rare-event discrepancy between Hbar_{β1,S} and Hbar_{β1} could in principle correlate with the large fluctuations of Hbar_{β2}.  This gap is repairable: Proposition 4.14 gives stretched-exponential tails for the quenched probability that the β1 polymer exits the strip S(ε,β2), and those tails are strong enough to imply E|Hbar_{β1,S} − Hbar_{β1}|² → 0, but this implication is not written in the paper. As written, the proof of the central independence claim relies on an unverified L² approximation.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The manuscript studies temperature chaos for the continuum directed random polymer (CDRP). Its central result, Theorem 1.7, states that when two CDRP free-energy profiles at inverse temperatures β1 and β2 are coupled through the same white noise and β1,β2→∞ with β1=o(β2), the pair converges in distribution to two independent directed landscapes. The proof is built on a thin-strip influence bound (Theorem 5.2), proved via a white-noise version of the Efron–Stein inequality, together with imported convergence and localization results for the CDRP and the directed landscape. As a byproduct, the authors prove that the directed landscape is a two-dimensional black noise (Theorem 1.9), resolving a conjecture of Virág. They also prove mutual singularity of CDRP polymer measures at different temperatures and formulate a conjecture for the sharp temperature-chaos exponent.","tokens_in":79160,"tokens_out":11239,"duration_ms":105399,"significance":"If completed, this would be the first rigorous energetic temperature-chaos result for a canonical KPZ-universality model, with the limiting pair explicitly identified as independent copies of the universal zero-temperature object. The measure-theoretic framework for restricted CDRP partition functions and the influence estimates are likely to be reusable. The black-noise theorem resolves an open conjecture and provides a third known example of a two-dimensional black noise. The paper is unusually careful about measurability and resampling of white noise, and the overall proof architecture is coherent and detailed.","major_comments":[{"comment":"Lemma 5.4 is invoked with X_n=Hbar^ξ_{β1}(u), Z_n=Hbar^ξ_{β1,S}(u), but the lemma requires limsup E||X_n−Z_n||^2 ≤ ε. Lemma 5.1 only establishes convergence in probability. Proposition 4.14 controls the quenched probability of polymer exit on a high-probability event, but it does not, as stated, control the size of |log p| on the exceptional event where the exit probability is not small. An additional integrability statement for the restricted free energy, or a modification of Lemma 5.4 requiring only convergence in probability, is needed. As written, the independence conclusion does not formally follow from the stated lemmas.","section":"§5, proof of Theorem 1.7 / Lemma 5.4"},{"comment":"There is a second formal mismatch with Lemma 5.4. The lemma requires ||Y_n−W_n||_p → 0 for the sequences associated with each ε>0. Theorem 5.2 gives only limsup E|Hbar^ξ_{β2}−Hbar^{η_S}_{β2}|^2 ≤ ε^α for a fixed ε>0. Since ε is fixed in the application, this bound does not vanish. To obtain the desired tensorization one must either diagonalize over ε→0 or state a variant of Lemma 5.4 in which both approximation errors satisfy limsup ≤ ε. The current proof does not supply this step, leaving the central claim incomplete.","section":"§5, proof of Theorem 1.7"}],"minor_comments":[{"comment":"The notation H^ξ_{β1,S} for the restricted free energy is very close to H^ξ_{β1}; consider a different symbol (e.g., H^{ξ,strip}_{β1}) to avoid confusion in a long proof.","section":"§5, Eq. (5.3)"},{"comment":"The endpoint scaling in the strip uses factors 2^{1/3}β^{2/3}; it would help to state explicitly how these factors are absorbed by the constants in Proposition 4.14 and the convergence in Theorem 1.5.","section":"§5, Lemma 5.1"},{"comment":"Section 7.9 is a sketch of a possible SHF result rather than a theorem. This is acceptable, but it should be labeled more prominently as heuristic/sketch so that it is not read as a claim.","section":"§1.5 / §7.9"},{"comment":"Some citations, e.g., [QR V25], contain a space in the author field. Please check the reference formatting.","section":"References"}],"recommendation":"major_revision","confidential_remarks":"The central proof has a localized but load-bearing gap in the application of Lemma 5.4: the hypotheses of that lemma are not verified, and one of them is not even the right type of statement. The gap appears repairable within the paper's scope, either by proving an L2 approximation or by replacing Lemma 5.4 with a weaker convergence-in-probability diagonal argument. I do not see grounds for rejection, but the manuscript cannot be accepted in its present form."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Here's my read after going through the visible portions (Sections 1–6 plus the statements of 7–9).\n\nThe paper is the real thing. Theorem 1.7 is the first rigorous energetic decoupling across temperatures for a KPZ-universality model, and Theorem 1.9 resolves Virág's two-dimensional black-noise conjecture for the directed landscape. Those are new, non-overlapping results, not repackaged versions of [Cha14] or [GH24a]. The proof architecture is serious: white-noise resampling, Wiener-chaos measurability for restricted partition functions, an Efron–Stein influence bound for thin strips, and careful completion of σ-algebras. I also credit the honest framing: β1 = o(β2) is far from the conjectured β^{1/3}, the ε^α exponent is weaker than the believed ε^{1/2}, and Sections 7.9 and 9 are labeled sketches.\n\nThe stress-test note lands. In the proof of Theorem 1.7, Lemma 5.4 is invoked with X_n − Z_n = H_{β1} − H_{β1,S}(u), but Lemma 5.1 only gives convergence in probability for finite-dimensional distributions. Lemma 5.4 requires an L^2 bound, and none is proved. The gap is almost certainly repairable using the stretched-exponential tails in Proposition 4.14 plus second-moment control, but as written the independence conclusion does not formally follow. That is the main concrete flaw I see.\n\nThe other soft spot is that the load-bearing influence estimate, Theorems 5.2/6.1, is exactly where the provided text truncates. I can see the edge bounds and the setup for the bulk bound, but I could not check the bulk bound in Section 6.5 or the completion of Theorem 5.2. The black-noise section is also mostly outside the visible text. This is not a demonstrated error, just an unverified load-bearing step.\n\nThe external inputs [Wu26] and [DZ24] are recent and are quoted without re-derivation. They are natural dependencies after [Wu26], but they make the proof hostage to their correctness. I do not see a circularity problem.\n\nVerdict: conditional. The reader's 6/10 soundness is about right. If the L^2 gap is closed and the bulk influence bound checks out, this is an important paper. Send it to referees; they should be asked to verify Lemma 5.4's hypotheses and the full proof of Theorem 6.1.","headline":"Genuinely new decoupling theorem and black-noise proof, with a repairable L2 gap in the independence argument and a load-bearing influence bound I could not fully verify in the visible text.","tokens_in":79901,"tokens_out":2689,"would_cite":true,"duration_ms":33236,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["60K35","60H15","82B44"],"pacs":[],"model":"deepseek-v4-flash","headline":"The paper proves that in a canonical random polymer model, free energies at two inverse temperatures become independent directed landscapes when the temperatures are far apart.","keywords":["temperature chaos","continuum directed random polymer","directed landscape","KPZ universality class","noise sensitivity","black noise","Gaussian multiplicative chaos","white noise resampling"],"falsifier":"Simulate two coupled CDRP free-energy profiles H_beta1 and H_beta2 with beta1=o(beta2), and measure the covariance of their finite-dimensional distributions; if the covariance does not vanish as beta2/beta1 →∞, Theorem 1.7 is false. A sharper check targets the localization input: measure the quenched transversal fluctuation exponent of the CDRP polymer at large beta; if the typical width is not of order beta^{2/3}, the strip argument fails.","tokens_in":78665,"feed_emoji":"🌡️","tokens_out":6339,"duration_ms":63890,"temperature":0.7,"pith_summary":"The paper aims to prove temperature chaos: a random system's optimal-energy landscape reorganizes completely when temperature is changed, not just slightly. It studies the continuum directed random polymer (CDRP), the canonical continuum model in the Kardar–Parisi–Zhang universality class, at two inverse temperatures beta1 and beta2 driven by the same white-noise environment. The main theorem states that when beta1 is much smaller than beta2 and both are large, the two free-energy profiles converge jointly to two independent copies of the directed landscape, the universal zero-temperature limit. This gives the first rigorous instance of energetic decoupling across temperatures in a KPZ model. A byproduct proves that the directed landscape itself is a two-dimensional black noise.","feed_headline":"Temperature chaos proved in a canonical polymer model","feed_subtitle":"Rigorous proof that the continuum directed random polymer forgets its low-temperature landscape as the temperature gap grows.","key_machinery":"The central mechanism is the thin-strip influence bound. For a spatial strip S of half-width epsilon beta2^{2/3}, the L2 sensitivity of the beta2 free energy to resampling the white noise inside S is at most epsilon^alpha for small epsilon. This is proved via an Efron–Stein inequality for white noise: the strip is tiled into boxes and the total influence is bounded by the sum of per-box influences. Each per-box influence is small because the polymer crosses a given small box with probability about the box's width in transversal units, and when it does, the energy change is controlled by the KPZ fluctuation on that box's scale. Brownian comparison estimates for the Airy line ensemble (or its","core_discovery":"Coupled through the same space-time white noise, the CDRP free energy profiles H_beta1 and H_beta2, scaled to their natural KPZ units, converge in distribution as beta1,beta2 →∞ with beta1=o(beta2) to (L1,L2), where L1 and L2 are independent directed landscapes. The theorem is proved by showing that the beta1-polymer is localized inside a thin spatial strip of width of order beta2^{2/3} around its typical path, that the beta1 free energy is nearly measurable with respect to the noise in that strip, and that resampling the noise inside any such thin strip changes the beta2 free energy negligibly. Thus the two free-energy profiles become functions of essentially independent pieces of the noise","pith_inferences":["One plausible extension is that the same thin-strip mechanism transfers from the CDRP to discrete lattice directed polymers in the strong-disorder regime, predicting the experimentally relevant n^{-1/6} temperature-perturbation scale; the paper only sketches this connection.","If the directed landscape is a two-dimensional black noise, then any finite-resolution white-noise approximation of the landscape sees it only through high-frequency components, a feature that could matter for numerical simulations of interface growth.","The same influence method may show that the critical two-dimensional stochastic heat flow is a three-dimensional black noise, which would be the first known example in dimension three or higher.","The relation between temperature and disorder chaos sketched in the paper suggests a testable equivalence: the critical temperature-perturbation scale should match the disorder-perturbation scale under a reparameterization, a prediction that finite-size simulations could probe."],"forward_implications":["If Theorem 1.7 holds, temperature chaos is rigorous in the CDRP: the free-energy profiles at well-separated temperatures are asymptotically independent, not merely decorrelated.","The same machinery proves the directed landscape is a two-dimensional black noise, making it the third known example after critical planar percolation and the Brownian web.","Polymer measures at any two distinct inverse temperatures are mutually singular almost surely, even when free energies are strongly correlated; the path-space Gibbs measures live on different energy-level sets.","For temperatures that are polynomially close, the coupled pair converges to two copies of the same directed landscape, so the chaos transition must occur at a nontrivial separation scale.","A sharp threshold is conjectured at beta2 - beta1 ~ beta1^{1/3}; the paper proves stability and decoupling on either side of that scale but leaves the threshold itself open."],"fun_headline_variants":["Polymers forget heat: temperature chaos proven","Temperature chaos: polymers scramble landscapes","Proof: Polymer paths decouple across temperatures","Heat scrambles polymer: independent landscapes emerge"],"cache_read_input_tokens":2304,"weakest_assumption_plain":"The argument stands on the CDRP polymer localization estimate imported from an earlier work: at inverse temperature beta and over a unit time interval, the polymer's transversal displacement has typical scale beta^{2/3} with stretched-exponential tails. If this scale is not exactly beta^{2/3}, the strip either fails to contain the beta1 polymer or the decoupling regime beta1=o(beta2) is not the correct one.","fun_headline_variants_meta":{"raw":{"variants":["Polymers forget heat: temperature chaos proven","Temperature chaos: polymers scramble landscapes","Proof: Polymer paths decouple across temperatures","Heat scrambles polymer: independent landscapes emerge"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000706,"raw_usage":{"total_tokens":3100,"prompt_tokens":909,"completion_tokens":2191,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":653,"completion_tokens_details":{"reasoning_tokens":2138}},"tokens_in":653,"tokens_out":2191,"duration_ms":16093,"temperature":1.0,"reasoning_tokens":2138,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-01T15:44:21.410582+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Simulate two coupled CDRP free-energy profiles H_beta1 and H_beta2 with beta1=o(beta2), and measure the covariance of their finite-dimensional distributions; if the covariance does not vanish as beta2/beta1 →∞, Theorem 1.7 is false. A sharper check targets the localization input: measure the quenched transversal fluctuation exponent of the CDRP polymer at large beta; if the typical width is not of order beta^{2/3}, the strip argument fails.","supporting_citations":[],"review_version":1}