{"id":"ab2dd1e3-9216-4774-b222-79d6f2c3317a","arxiv_id":"2506.14071","paper_version":1,"verdict":"REJECT","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"high","formal_verification":"none","parameter_count":0,"one_line_summary":"The paper proposes a weighted-sum revision of the quantum extremal surface formula, but the weight is assumed rather than derived.","lead":"This paper proposes a new formula for the entropy of an evaporating black hole that averages over many surfaces instead of picking one. The derivation is not complete, so the proposal remains an unproven revision to a standard method.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The weight in the revised entropy formula is introduced by replacing the on-shell action of Σ_w with its entanglement entropy (Eq. 4.2 → 4.3); this identification is not derived and is not generally true, so Eq. (4.6) has no basis.","rationale":"The reader's weakest_assumption is exactly the same load-bearing concern I identify: Eq. (4.3) rests on an unproven identification of the on-shell action of Σ_w with its entanglement entropy. My reading of the paper confirms that no derivation of this replacement is given; the text moves directly from the path-integral expression in Eq. (4.2) to the entropy-weighted expression in Eq. (4.3) with only the qualitative phrase quoted above. This matters because every subsequent formula—the weighted average in Eq. (4.4), the continuum version in Eq. (4.6), and the claimed reduction to the QES result in Eq. (4.7)—inherits this step. Without Eq. (4.3), the paper offers a speculative proposal, not a derivation. I do not see independent support that would rescue the claim: there is no formal verification, reproducible code, or parameter-free derivation, and the paper itself concedes that a more detailed analysis is required. The validation in Section 5 also assumes the shape of S_semi-cl(Σ_X) to produce the desired early and late behavior, so it does not test the central formula independently. I therefore agree with the reader's reject verdict, and my stress-test does not change it. I note in passing that the Section 3 critique of the saddle-point approximation is also debatable, since replica-trick computations are normally performed at integer n>1 with analytic continuation of the result, but I do not need to rely on that secondary concern to identify the load-bearing gap.","tokens_in":8453,"tokens_out":6970,"duration_ms":80859,"concrete_test":"Compute both \\tilde I(Σ_w) and S_semi-cl(Σ_w) explicitly in the JT-gravity-plus-CFT model of Refs. [18,19] for a family of twist-operator positions w. For a concrete realization, take a free-fermion CFT on an interval [w, L_cutoff]; S(Σ_w) is known exactly from the CFT result, while \\tilde I(Σ_w) is the on-shell Euclidean action of the same interval. If exp(-n \\tilde I(Σ_w)) differs from exp(-n S(Σ_w)) by more than a constant independent of both n and w, Eq. (4.3) is false and the central formula Eq. (4.6) collapses. An even more direct check is to compare Eq. (4.3) with the exact replica computation of tr ρ^n in the same model.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The entire derivation of the revised entropy formula hinges on one step: between Eqs. (4.2) and (4.3), the authors replace exp(-n \\tilde I(Σ_w)) by exp(-n S(Σ_w)), with the justification that the on-shell action of region Σ_w 'characterizes the degrees of freedom contained in Σ_w, allowing us to replace it with the entanglement entropy of the fields on Σ_w.' This is not a derivation. The on-shell Euclidean action of a region is a bulk functional determined by the classical solution, while the von Neumann entropy S(Σ_w) is a different, generally non-local functional of the state; in a 2D CFT, for instance, S(Σ_w) grows logarithmically with interval length while the local action is extensive. Nothing in the replicated path integral presented in Section 4 shows that the n-dependent exponent receives the coefficient -n S(Σ_w) rather than -n \\tilde I(Σ_w). If this identification fails, the weighted sum, the log term, and the reduction to Eq. (4.7) are all unsupported. The paper itself flags the gap by stating immediately after introducing the weight that 'a more detailed analysis was required to demonstrate this.' The algebra from Eq. (4.3) to (4.4) is consistent, but the starting point is postulated, not derived. The alternative concern that the n→1 saddle-point critique in Section 3 is itself questionable adds to the risk, but the action-entropy identification is the load-bearing gap.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"This paper revisits the quantum extremal surface (QES) prescription for black hole entanglement entropy. It argues that the standard replica-trick derivation of the QES formula relies on a saddle-point approximation in the number of replicas n that becomes invalid in the limit n → 1, because the prefactor (n−1) suppresses the extremization condition. To address this, the paper proposes a revised entropy formula, Eq. (4.6), which replaces the min-extremum over a single surface by a weighted sum over all surfaces X from 0 to rSch, with weight exp(−S_semi-cl(Σ_X)) plus a logarithmic term. The authors claim that this formula reduces to the generalized entropy at an extremal surface under a sharply-peaked approximation, and that it reproduces the Page curve in early and late evaporation stages.","tokens_in":8760,"tokens_out":5372,"duration_ms":55640,"significance":"If the proposed revision were rigorously derived, it would constitute a significant modification of the QES prescription, altering the island formula and the way extremal surfaces are selected in semiclassical gravity. The paper correctly emphasizes the subtlety of the n → 1 limit in the replica trick, a point that is often glossed over. However, the central new step—replacing the on-shell action of a region by its entanglement entropy—is asserted rather than derived, and the paper itself acknowledges that 'a more detailed analysis was required.' The validation in Section 5 is also circular: it assumes the behavior of S_semi-cl(Σ_X) needed to obtain the Page curve rather than deriving it from an independent calculation. The paper contains no machine-checked proofs or concrete model computations; its contribution is a plausible heuristic proposal, not a supported derivation. For these reasons, the significance of the result is currently outweighed by the unsupported load-bearing step.","major_comments":[{"comment":"The transition from exp(−n \\tilde I(Σ_w)) to exp(−n S(Σ_w)) is the foundation of the entire revised formula, yet it is not justified. The on-shell Euclidean action of a region is a local functional determined by the classical solution, whereas the entanglement entropy S(Σ_w) is a generally non-local functional of the quantum state; there is no general identity equating them. The paper's justification—that the action 'characterizes the degrees of freedom'—is qualitative and does not follow from the replicated path integral presented in Section 4. The authors themselves concede that 'a more detailed analysis was required to demonstrate this.' Since the weight in Eq. (4.6) arises solely from this identification, the central claim is unsupported without this step.","section":"Section 4, between Eqs. (4.2) and (4.3)"},{"comment":"The decomposition of the replicated path integral into region (i) and region (ii) with the factorized form exp(−(n−1)Sgen(w)) ∫_{region (ii)} exp(−n I) is not derived. The coefficients (n−1) and n are read off from the off-shell action in Eq. (2.3), but it is not shown that the path integral of the full replicated system separates into a product of contributions from the two regions with these weights. In particular, the treatment of the twist operator and the boundary conditions at the gluing surface is not specified, and no concrete model is used to verify the factorization. Without such a derivation, Eq. (4.1) is an assumption about the path integral structure.","section":"Section 4, Eq. (4.1)"},{"comment":"The validation of the revised formula is circular. At early times, the result S ≈ Sgen(X=0) follows from the assumed linear growth S_semi-cl(Σ_X) ≈ O(N X/rSch), and at late times the result S ≈ Area(X=rSch)/4G follows from the assumed suppression of all X ≠ rSch. These are not independent predictions of Eq. (4.6) but rather properties inserted into S_semi-cl to obtain the Page curve. The paper does not compute S_semi-cl from a specific state or model, nor does it compare the weighted sum against an independent calculation. Consequently, the agreement with the Page curve is an artifact of the assumptions, not evidence for the formula.","section":"Section 5, Eq. (5.1) and Eq. (5.3)"},{"comment":"The reduction to the generalized entropy at the extremal surface \tilde X, Eq. (4.7), assumes that the probability distribution exp(−S_semi-cl(Σ_X)) is sharply peaked. However, Section 3 argues that the analogous distribution exp(−(n−1)Sgen(w)) is not sharply peaked because the prefactor (n−1) vanishes as n → 1. The paper does not explain why the peak in exp(−S_semi-cl) is justified when no small parameter multiplies S_semi-cl in the exponent; the distinction is not self-evident and requires a separate argument beyond the heuristic that S_semi-cl is large.","section":"Section 4, Eq. (4.7) and Section 3"}],"minor_comments":[{"comment":"The expression for the derivative of the Rényi entropy near n = 1 is written as d/dn Sn|_{n→1} = S/(2(n−1))|_{n→1} → ∞, which is ambiguous: the divergence is not literally S/(2(n−1)) but arises from the second-order expansion of Tr ρ^n; the notation should be clarified.","section":"Section 3, Eq. (1.4)"},{"comment":"The phrase 'Let check whether' should be 'Let us check whether'.","section":"Section 5, first sentence"},{"comment":"The caption reads 'Left is what Hawking considered in his calculations' which is informal; it should say 'The left panel shows the Hawking saddle'.","section":"Section 2, Fig. 2 caption"},{"comment":"The subscript w in ⟨Sgen(w)⟩_w and ⟨S(Σ_w)⟩_w is reused for the summation variable, which can be confusing; using a different index (e.g., j) for the expectation value would improve readability.","section":"Section 4, Eqs. (4.4)–(4.5)"},{"comment":"The papers [18,19] contain extensive discussion of the n → 1 limit and the factorization of the replica path integral; the manuscript would benefit from engaging with those explicit analyses rather than only citing them as the source of the off-shell action.","section":"References"}],"recommendation":"reject","confidential_remarks":"To the editor: the paper identifies a real subtlety in the n → 1 limit of the replica trick, but the proposed resolution is not derived. The key identification between on-shell action and entanglement entropy is explicitly left for future work by the authors. Given that the central formula rests on this unproven step and the validation is circular, I do not see how the result can be accepted in its current form. A resubmission that derives the weight from a concrete model (e.g., JT gravity with a full replica path integral) could change this assessment."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The paper identifies a real subtlety in the replica derivation of the QES prescription: the twist operator equation of motion is multiplied by (n−1), so the saddle point becomes non-selective as n→1. That observation is worth taking seriously. What is new is the proposed fix—a weighted sum over surfaces with weight exp(−S_semi-cl(Σ_X)), plus a log term—and the algebra from Eq (4.3) to (4.6) is internally consistent. The mutual information interpretation in Eq (4.10) is also a nice touch.\n\nThe problem is that the central step is postulated, not derived. Between Eqs (4.2) and (4.3), the on-shell action of region Σ_w is replaced by the entanglement entropy S(Σ_w) with the hand-wave that the action 'characterizes the degrees of freedom.' That identification is not generally true: the bulk action is extensive, while the entanglement entropy is non-local and in 2D CFT grows logarithmically. The paper itself concedes, immediately after, that 'a more detailed analysis was required to demonstrate this.' That is a red flag: the entire weighted-sum formula rests on this step, so without a derivation Eq (4.6) is unsupported.\n\nThe validation in Section 5 does not rescue it. The early and late time peaks follow from assumed behavior of S_semi-cl(Σ_X)—linear growth in X before the Page time, huge for X<r_Sch after—so the Page curve is put in by hand via the weight. And in the sharp-peak limit the formula reduces to the standard QES result, so the practical impact outside that limit is unclear. The claim that the extremal surface changes to one extremizing S_semi-cl rather than S_gen is also stated without support.\n\nThis is not a cranky paper. It engages the right literature, is transparent about its gaps, and the underlying question—whether the n→1 limit of the replica trick invalidates the saddle point—is legitimate and under-explored. But as it stands, the central formula is a conjecture. I would send it to a serious referee, because the issue deserves airing, but the referee should ask for the action-entropy identification to be derived or the paper to be reframed as a proposal with a proof-of-principle example. As is, it should not be accepted.","headline":"A genuine n→1 subtlety in the QES replica derivation, but the proposed weighted-sum fix is postulated rather than derived.","tokens_in":9284,"tokens_out":3265,"would_cite":false,"duration_ms":33277,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"The paper argues that the entanglement entropy of an evaporating black hole is a weighted average over all candidate surfaces, not the single extremal surface picked by the QES prescription.","keywords":["entanglement entropy","quantum extremal surface","black hole evaporation","replica trick","saddle point approximation","generalized entropy","island","unitarity curve"],"falsifier":"Compute the full replicated path integral for a solvable two-dimensional model of an evaporating black hole without replacing the on-shell action of $\\Sigma_w$ by its entanglement entropy, and take $n\\to 1$; if the result differs from Eq. (4.6), the weighting is wrong.","tokens_in":8158,"feed_emoji":"🕳️","tokens_out":10713,"duration_ms":94537,"temperature":0.7,"pith_summary":"The paper argues that the standard quantum extremal surface (QES) prescription for the entanglement entropy of an evaporating black hole is incomplete. In the replica-trick derivation, the term that selects the extremal surface is proportional to $n-1$ and vanishes as the number of replicas $n$ approaches one, so the saddle-point approximation that justifies a single surface breaks down. The authors derive a revised formula in which the entropy is a weighted average over all candidate surfaces $X$ inside the black hole radius, with weight $\\exp(-S_{\\rm semi\\text{-}cl}(\\Sigma_X))$, plus a logarithmic term. Early and late in evaporation this formula reproduces the expected unitary behavior, and when the weight is sharply peaked it reduces to the generalized entropy at an extremal surface. If correct, it changes the central object of the calculation from one special surface to an entire distribution of surfaces.","feed_headline":"Black hole entropy is a weighted sum, not one extremal surface","feed_subtitle":"A revised formula reproduces unitary evaporation and reduces to the old extremal-surface rule only as an approximation.","key_machinery":"The central object is a weighted ensemble of codimension-2 surfaces: every surface $X$ between the origin and the black hole radius is assigned weight $\\exp(-S_{\\rm semi\\text{-}cl}(\\Sigma_X))$, where $\\Sigma_X$ is the region between $X$ and the cutoff surface. The weight is generated by evaluating the on-shell action of the degrees of freedom outside the gluing surface, which carry a factor $n$ in the replicated action; the location of the gluing surface itself is governed by a term of order $n-1$, so no saddle point selects it and the path integral sums over all locations. The same mechanism explains why a naive saddle-point treatment fails: the width of the integrand in the surface location diverges as $n\\to 1$ unless a cutoff is introduced, and the divergent derivative of the $n$-dependent replica entropy at $n=1$ forbids a finite cutoff.","core_discovery":"The central claim is Eq. (4.6): the fine-grained entropy of the black hole should be $$S = \\frac{\\sum_{X\\le r_{\\rm Sch}} [S_{\\rm gen}(X)+S_{\\rm semi\\text{-}cl}(\\Sigma_X)] $e^{{-S_{\\rm semi\\text{-}}$cl}(\\Sigma_X)}}{\\sum_{X\\le r_{\\rm Sch}} $e^{{-S_{\\rm semi\\text{-}}$cl}(\\Sigma_X)}} + \\ln \\sum_{X\\le r_{\\rm Sch}} $e^{{-S_{\\rm semi\\text{-}}$cl}(\\Sigma_X)},$$ not the single extremal surface selected by the minimality rule. The weight emerges from the path integral over the region outside each gluing surface, evaluated on shell, and is not visible from the replicated action alone. In the early stages, the rapidly growing semi-classical entropy makes only nearly vanishing surfaces contribute, giving the increasing entropy of the early semi-classical saddle; at late times, only surfaces near the black hole radius contribute, giving the decreasing area term required by unitarity. When the weight is sharply peaked at a particular surface, the formula reduces to the generalized entropy at that surface, recovering the QES answer with a different selection rule.","pith_inferences":["If Eq. (4.6) is correct, the QES answer is only the leading term of an expansion; the first corrections are controlled by the spread of $S_{\\rm semi\\text{-}cl}(\\Sigma_X)$ and should be computable in tractable models.","A numerical replica calculation in a toy model would test the on-shell identification directly, since the exact integral over gluing-surface locations is in principle well defined.","The 'correlations cost' interpretation suggests a variational principle for the island region that could be applied to settings beyond the one considered in the paper."],"forward_implications":["Early in evaporation the weight is dominated by surfaces near $X=0$, so the entropy grows as the semi-classical entropy of the interior.","Late in evaporation only surfaces near $X=r_{\\rm Sch}$ contribute, yielding the decreasing area that makes the evolution unitary.","When the weight is sharply peaked, the formula reduces to the generalized entropy at an extremal surface, so the QES prescription survives as a special case.","Higher replica contributions are expected to change the value of the entropy but leave the weight $\\exp(-S_{\\rm semi\\text{-}cl}(\\Sigma_X))$ unchanged.","The weight's information-theoretic form implies that the dominant surface maximizes the number of enclosed microstates while minimizing the correlations it cuts across."],"supporting_citations":[{"why":"Defines the quantum extremal surface prescription and the minimality rule that the paper revises.","marker":"[7]"},{"why":"Derives the QES prescription from replica wormholes and supplies the replicated action Eq. (2.3) whose n-to-1 limit is under examination.","marker":"[18]"},{"why":"Provides an independent replica-wormhole derivation of the QES prescription used as a reference point.","marker":"[19]"},{"why":"Documents the failure of analytic continuation and the divergent derivative of the replica entropy at n-to-1, motivating the revision.","marker":"[16]"},{"why":"Establishes the unitary time dependence of black hole radiation entropy that the revised formula must reproduce.","marker":"[9–11]"},{"why":"Defines the island region whose connectivity explains why the extremal surface is selected by the weight.","marker":"[22–25]"},{"why":"Supports the claim that spacetime connectivity and mutual information are linked, used to interpret the weight.","marker":"[26, 27]"}],"fun_headline_variants":["Quantum extremal surfaces get a weighted-sum upgrade","Black hole entropy: weighted sums replace single extremum","Entropy from many surfaces, not just the extremal one","Revised black hole entropy: weighted sum over surfaces","Black hole entropy: many surfaces, not one extremal"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The calculation assumes, without deriving it from the path integral, that the on-shell action of the region inside each trial surface equals the entanglement entropy of the fields in that region, and if that equality fails the weighted formula collapses.","fun_headline_variants_meta":{"raw":{"variants":["Quantum extremal surfaces get a weighted-sum upgrade","Black hole entropy: weighted sums replace single extremum","Entropy from many surfaces, not just the extremal one","Revised black hole entropy: weighted sum over surfaces","Black hole entropy: many surfaces, not one extremal"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000657,"raw_usage":{"total_tokens":2984,"prompt_tokens":900,"completion_tokens":2084,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":516,"completion_tokens_details":{"reasoning_tokens":2007}},"tokens_in":516,"tokens_out":2084,"duration_ms":15690,"temperature":1.0,"reasoning_tokens":2007,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-07T00:18:21.272426+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Compute the full replicated path integral for a solvable two-dimensional model of an evaporating black hole without replacing the on-shell action of $\\Sigma_w$ by its entanglement entropy, and take $n\\to 1$; if the result differs from Eq. (4.6), the weighting is wrong.","supporting_citations":[],"review_version":1}