{"id":"e492ec9e-ff5b-45cb-b90c-7210c8a0d30a","arxiv_id":"2412.15064","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":6,"one_line_summary":"Non-local hybrid metric-Palatini gravity generically contains ghosts, but a degenerate subclass with metric f(R) gravity plus Palatini non-local terms is ghost-free and can drive slow-roll inflation resembling Starobinsky inflation with a spectator field.","lead":"This paper studies a modified theory of gravity that mixes standard and Palatini curvature with non-local terms built from inverse d'Alembert operators. It finds that the general theory has ghost instabilities, but a special hybrid version can support inflation and behaves like single-field inflation with a spectator field.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Localization of the non-local hybrid action into a scalar-tensor theory is assumed, not demonstrated; if the auxiliary fields carry spurious homogeneous modes, the ghost count and no-ghost conditions of Secs. 2-3 do not constrain the original non-local theory.","rationale":"The reader's weakest_assumption identifies exactly the same load-bearing point: the dynamical equivalence between the non-local action and the local auxiliary-field representation. I find this concern justified and worth making more explicit. The algebra in Secs. 2 and 3 is carefully done; the ghost-counting for non-degenerate models (N=n+m ghosts from the pairs in Eq. (2.21)) follows from the structure of the kinetic matrices, and the Sylvester inequalities for L1 and L2 are internally consistent. The weak spot is the foundational localization step: the paper cites Ref. [115] and states that localization 'bypasses' the boundary-condition ambiguity, but a local theory with free initial data for the auxiliary fields cannot be equivalent to a non-local theory with a fixed Green's function unless the homogeneous modes are constrained. This matters most for the degenerate models, where the standard Legendre inversion fails and the auxiliary sector is introduced by hand. The proposed test (direct propagation analysis of L2) would settle the issue. The paper also explicitly limits its inflation analysis to background dynamics and hand-picked parameters in the Conclusions, so the CONDITIONAL verdict is appropriate; my analysis does not move it.","tokens_in":22461,"tokens_out":36548,"duration_ms":293123,"concrete_test":"Take model L2 of Eq. (3.1) with f(R)=a_2 R+b_2 R^2, G(\\beta)=g\\beta, V=0. Derive the second-order action around Minkowski space directly from the non-local action, representing \\Box^{-1} as the retarded Green's function, and count the number of propagating scalar degrees of freedom from the principal symbol. Repeat for the localized action (3.7) after enforcing \\Box\\beta=R. If the numbers or pole structures differ, the localization introduces spurious modes and the no-ghost inequalities of Sec. 3 do not pertain to the original theory.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The paper's central no-ghost result rests on replacing the non-local action (2.1) by the local scalar-tensor action (2.4)-(2.6) with auxiliary fields alpha_i, beta_j and Lagrange multipliers lambda_i, rho_j. This localization is not a standard Legendre transform for the degenerate models of Sec. 3: the Hessian condition F_{\\chi\\chi}F_{\\eta\\eta}-F_{\\chi\\eta}^2\\neq0 is explicitly violated, so the procedure is a formal rewriting whose dynamical equivalence is not proven. The localized theory contains homogeneous solutions of \\Box\\alpha_i=R and \\Box\\beta_j=R as free initial data; in a retarded-Green's-function formulation of \\Box^{-1} these modes are absent. The ghost pairs exhibited after Eq. (2.21) and the Sylvester inequalities (e.g., G'(\\beta)>-\\psi/6 for L2 in Eq. (3.8)) are properties of the localized field space, and would not apply if the auxiliary-sector degrees of freedom are spurious. Since the rest of the paper (including inflation in Sec. 4) is built on this scalar-tensor representation, the equivalence is the most load-bearing assumption in the argument.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"This paper studies non-local extensions of hybrid metric-Palatini gravity in which inverse d'Alembert operators act on both the metric Ricci scalar R and the Palatini Ricci scalar R̄. Following Ref. [115], the authors replace the non-local action (2.1) by a local scalar-tensor action with auxiliary fields and Lagrange multipliers, Eqs. (2.4)-(2.6). For non-degenerate F they show that the Einstein-frame kinetic structure contains at least m + n ghosts, where m and n are the highest powers of the inverse d'Alembert operator acting on the two curvatures. They then study degenerate models in which the Hessian condition F_χχ F_ηη − F_χη² ≠ 0 is violated: L1 (Palatini f(R) with metric non-localities) and L2 (metric f(R) with Palatini non-localities). Applying Sylvester's criterion to the resulting 3×3 kinetic matrices, they derive algebraic no-ghost conditions, e.g. G'(β) > −ψ/6 for L2. The second half is a numerical study of slow-roll inflation in the Einstein frame for quadratic f, with power-law and exponential kinetic couplings and with and without a potential V(□^{-1}R). The main claims are that non-degenerate hybrid non-local actions are generically ghost-ridden, that the degenerate models restore stability, and that L2 supports slow-roll inflation, effectively reducing to single-field inflation with a spectator field.","tokens_in":22697,"tokens_out":31853,"duration_ms":211615,"significance":"If correct, the central ghost-counting result — that the number of ghosts equals the sum of the highest powers of □^{-1} acting on the metric and Palatini curvatures — cleanly extends the De Felice-Sasaki analysis of Ref. [115] to the hybrid metric-Palatini setting and rules out a broad class of non-local hybrid actions, isolating the degenerate models of Sec. 3 as the only stable candidates. The algebraic derivations in Secs. 2 and 3 are coherent and transparent, and the explicit no-ghost inequalities (the Sylvester conditions after Eqs. (3.3) and (3.8), and the field-space form (4.9)) are concrete and checkable. The identification of the L2 model as a Starobinsky-like plateau theory with an effectively frozen spectator field is a useful first step toward phenomenology, and the paper is honest about its background-only scope. Its main weaknesses are the undocumented numerical support for the inflationary claim (no e-fold counts, initial conditions, or sensitivity analysis) and the assumed rather than proven localization equivalence. No code is provided, but the results are in principle reproducible from the stated equations.","major_comments":[{"comment":"The central ghost-counting and no-ghost results are derived in the localized scalar-tensor representation, and the paper explicitly adopts the 'perspective of considering the non-local theory as equivalent to a local scalar-tensor model' (Sec. 1). The dynamical equivalence is assumed, not demonstrated, and the degenerate models of Sec. 3 are precisely the cases in which the Legendre inversion condition F_χχ F_ηη − F_χη² ≠ 0 of Eq. (2.7) fails: the localization of L1 and L2 is a formal rewriting whose equivalence to the original non-local action is not established. The auxiliary fields α_1, β_1 obey □α_1 = R and □β_1 = R̄ and carry homogeneous solutions of the wave operator as free initial data, which are absent if □^{-1} is defined with a retarded Green's function. The ghost pairs exhibited after Eq. (2.21) and the Sylvester inequalities (e.g. G'(β) > −ψ/6 for L2, after Eq. (3.8)) are therefore, strictly speaking, statements about the localized field space. Since the inflationary analysis of Sec. 4 is built entirely on this representation, the paper should either justify the equivalence for the degenerate case (or cite a result covering it) or state explicitly in the abstract and in Sec. 3 that the ghost count refers to the localized definition of the theory.","section":"Secs. 2-3, Eqs. (2.4)-(2.6), (2.7), (3.1)"},{"comment":"The paper's central inflationary claim — that the L2 model can support slow-roll inflation with 'the adequate number of e-folds' and that the no-ghost and slow-roll conditions are checked 'a posteriori' along the evolution (abstract and Sec. 5) — is not quantitatively documented in the body. No e-fold number is reported for any of the cases in Figs. 1-4, the initial conditions of the integrations are not given, and the evolution of the diagnostic quantities (the no-ghost inequality (4.9) and ϵ0 = −Ḣ/H²) is not shown. As printed, the reader cannot verify that inflation lasts for, say, 50-60 e-folds, or that the trajectory stays inside the no-ghost region. I request a table reporting N_e, the initial field values, and the maximum of |(dχ2/dΨ)Ξ/6| along each trajectory for the cases displayed in Figs. 1-4.","section":"Sec. 4.1-4.2, Figs. 1-4"},{"comment":"The text states that 'for σ1 = 1 the potential Y1 is independent of the field Φ', but substituting σ1 = 1 into Eq. (4.11) gives Y1 = (Ξ²/6 + a1)² / (4b1 e^{2√(2/3)Φ}), which depends on Φ through the exponential factor; the same Φ-dependence follows from Eqs. (3.5)-(3.6), where the potential appears as W1(Ψc, Ξc)/e^{2√(2/3)Φc}. The stated reason for excluding the L1 model from the inflationary analysis ('infinite slow-rolling stage along one scalar field direction', Sec. 5) therefore rests on an incorrect premise. The argument must be corrected; the conclusion may survive, since Y1 has no minimum in the Φ-direction, but the reasoning as printed is invalid.","section":"Sec. 4, text after Eq. (4.11)"},{"comment":"The inflationary results are obtained with hand-picked parameters a2 = 2.3, b2 = 0.001, k = 0.1 and no sensitivity analysis. The text asserts that 'the shape of the potential is not tightly constrained by these chosen values' (Sec. 4.1) without reporting any scan over a2, b2, or k, although these parameters fix the location and height of the minimum; for the quoted values the plateau height is ~a2²/(4b2) ≈ 1.3×10³ in Planck units, and no statement is made about the implied scalar amplitude A_s (which the paper defers to future perturbation analysis). Given that the abstract advertises how the kinetic couplings 'influence the number of e-folds', a sensitivity scan over (a2, b2, k, V0) and over initial Φ is needed to substantiate the feasibility claim beyond a single point in parameter space.","section":"Sec. 4.1-4.2, Eqs. (4.10)-(4.22)"},{"comment":"The abstract and the conclusions claim that the paper 'assessed the well-posedness of the first-order slow-roll parameter, which ultimately resulted in additional constraints among the derivatives of the potential and the fields'. In the body, the only slow-roll criterion stated is the positivity of ϵ0 ≡ −Ḣ/H² after Eq. (4.8), and no constraints on potential derivatives are derived anywhere in Sec. 4. Either the promised derivation should be included, or the claim should be amended to match the content.","section":"Abstract and Sec. 5 vs. Sec. 4"}],"minor_comments":[{"comment":"The two Lagrangians L1 and L2 are typeset identically in Eq. (3.1), since the overbars distinguishing the Palatini curvature from the metric curvature are not rendered; the distinction must be clearly visible in the published version, as the entire degeneracy argument relies on the two curvatures being different.","section":"Eq. (3.1)"},{"comment":"The phrase 'condition FRRFRR − F²RR ≠ 0 is now evaded' should say that the non-degeneracy condition is violated (the models are degenerate), and the notation should carry the overbars; 'evaded' is misleading.","section":"Sec. 3, second paragraph"},{"comment":"The power-law formula (C.1) is not valid at n = 0, which is one of the cases used in Figs. 1 and 3; the logarithmic limiting expression should be given explicitly.","section":"App. C, Eq. (C.1)"},{"comment":"The initial conditions, integration domains, and numerical tolerances for the integrations shown in Figs. 1-4 are not stated, and the meaning of 'normalised' in the figure captions (normalised to what quantity?) is not defined.","section":"Sec. 4, Figs. 1-4"},{"comment":"The text describes k = 0.1 both as a small parameter making the non-local terms act 'as perturbations' and (for n = 1) as making the non-local coupling 'significant'; the intended hierarchy between these two statements should be clarified.","section":"Sec. 4.1.1"},{"comment":"The units of the numerical parameters (a2, b2, k, V0) are never stated; the paper should specify that they are in reduced Planck mass units (or equivalent).","section":"Sec. 4"},{"comment":"The light-field condition is phrased as 'H²/YΨΨ ≫ 1'; stating it as YΨΨ/H² ≪ 1 would make the connection with the standard m² ≪ H² criterion for light fields clearer.","section":"Sec. 4.2.1"},{"comment":"The derivation of the field-space no-ghost inequality (4.9) from the Sylvester condition G'(β) > −ψ/6 of Sec. 3, which uses the field redefinitions (3.4) and the identification ψ = −Ξ²/6, should be displayed, since (4.9) is the form used in all the numerical checks.","section":"Sec. 4, Eq. (4.9)"},{"comment":"There is a typo in the sentence 'Variation of Eq. (2.1) with respect to λ_i, ρ_j guarantees that the original formulation is consistently recovered, how it is showed by'; it should read 'as shown by'.","section":"Sec. 2, after Eq. (2.4)"}],"recommendation":"major_revision","confidential_remarks":"This is a competent, well-written extension of Ref. [115] to hybrid metric-Palatini gravity, and the ghost-counting results appear correct under the stated localization assumption. My main concerns are that the inflationary part, advertised in the abstract as a central result, is quantitatively under-documented (no e-fold numbers, no initial conditions, no sensitivity analysis), and that there is an internal inconsistency in the treatment of the L1 model (the claimed Φ-independence of Y1 contradicts the paper's own Eq. (4.11)). The localization caveat applies with equal force to the ghost-counting claims, though the authors are transparent about following Ref. [115]. With an e-folds table, stated initial conditions, a sensitivity scan, and a correction of the Y1 statement, the paper would be publishable; the numerical section currently reads as work in progress, and I would not accept it in its present quantitative state."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Short version: the paper is a legitimate extension of De Felice–Sasaki ghost counting to hybrid metric-Palatini gravity, and the ghost-free L2 model (metric f(R) plus Palatini non-local terms) is a genuinely new construction. The inflationary application is plausible but backed only by background numerics with hand-picked parameters, and the localization of the non-local action into a scalar-tensor theory is assumed rather than proven for the degenerate cases.\n\nThe core result in Sec. 2 is clean: for non-degenerate F(R,R,□^{-1}R,...), the number of ghosts equals the sum of the highest powers acting on metric and Palatini curvature. The derivation is algebraic and matches the structure of Ref [115]. The Sylvester-criterion analysis for L1 and L2 is straightforward and yields sensible no-ghost inequalities. The L2 model, with a Starobinsky-like plateau potential in the Einstein frame, is worth having as a concrete example.\n\nThe main soft spot is the one the stress-test note flags: the transition from (2.1) to (2.4)–(2.6) is a formal localization, not a proven equivalence. For the degenerate models in Sec. 3 the Hessian condition is explicitly violated, so the Legendre transform is suspect. The auxiliary fields introduce homogeneous solutions of □α=R that would be absent in a retarded-Green's-function formulation. If those modes are spurious, the ghost count and the no-ghost inequalities (3.8) apply to the localized field space, not to the original action. The authors never address this. I don't think it's fatal—linear coupling to curvature is the case where localization is most defensible—but a referee should ask for a derivation or at least a careful statement of what the localization defines.\n\nThe inflation section is the weak third act. The parameters a2=2.3, b2=0.001, k=0.1 are chosen by hand, there is no sensitivity analysis, no e-fold table, no perturbation spectra. The authors openly admit the background-only limitation. For a phenomenological claim, that's thin. The ghost result stands on its own; the inflation part is a preview.\n\nBottom line: this deserves a serious referee. The ghost-counting analysis is a genuine contribution, and the L2 model is a useful testbed. The referee should press on the localization issue and on more complete inflationary numerics, but the paper is not a desk reject.\n\nReading group: maybe—the ghost-counting part is worth discussing with students, but the inflation part would need better support first.\n\nI would cite it if I worked on non-local or hybrid gravity; the ghost-free example is a useful reference point.","headline":"A solid extension of non-local ghost counting to hybrid metric-Palatini gravity, with a genuinely new ghost-free model, but the inflationary claims need stronger support and the localization step deserves scrutiny.","tokens_in":23307,"tokens_out":2887,"would_cite":true,"duration_ms":25759,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":["04.50.Kd","98.80.Cq"],"model":"deepseek-v4-flash","headline":"Non-local hybrid gravity is ghost-ridden unless actions are degenerate.","keywords":["non-local gravity","hybrid metric-Palatini gravity","ghost instabilities","scalar-tensor equivalence","inflation","inverse d'Alembert operator","Starobinsky inflation","multi-field cosmology"],"falsifier":"Take the simplest non-degenerate case $m=n=1$ in action (2.1) and perform a Hamiltonian or Ostrogradski constraint analysis directly on the non-local action, without localizing $\\Box^{-1}$; if the ghost number is not $2$, or if it changes when boundary conditions of the inverse d'Alembert operator are varied, the stability claim fails. A second decisive test is to exhibit any non-degenerate hybrid action of the form (2.1) that is ghost-free, which would directly contradict the claimed unavoidability of $m+n$ ghosts.","tokens_in":22206,"feed_emoji":"🌀","tokens_out":14188,"duration_ms":100174,"temperature":0.7,"pith_summary":"The paper tries to establish that adding inverse-d'Alembert non-localities, of the form $\\Box^{-k}R$ and $\\Box^{-k}\\mathcal{R}$, to hybrid metric-Palatini gravity generically introduces ghost instabilities, and that only a degenerate subclass of actions is dynamically stable. It shows that in the local scalar-tensor version of a general non-degenerate action $F(R,\\mathcal{R},\\ldots,\\Box^{-m}R,\\ldots,\\Box^{-n}\\mathcal{R})$, the number of ghost fields is set by the sum $m+n$ of the highest non-locality powers on the two curvatures. It then constructs two ghost-free hybrid models, a Palatini $f(\\mathcal{R})$ action with metric non-localities and a metric $f(R)$ action with Palatini non-localities, and derives algebraic inequalities on the fields and the non-local coupling that guarantee positive-definite kinetic matrices. The paper's main cosmological result is that the metric-$f(R)$ plus Palatini-non-local case supports a finite slow-roll inflationary phase in the Einstein frame, reducing effectively to single-field inflation with a spectator field, while the Palatini-$f(\\mathcal{R})$ plus metric-non-local case has a flat direction and cannot end inflation. This matters because it narrows the space of non-local hybrid gravity models that can describe the early universe and ties their viability to the linear-coupling, degenerate form of the action.","feed_headline":"Non-local hybrid gravity is ghost-ridden unless actions are degenerate","feed_subtitle":"Adding inverse-d'Alembert terms to both curvatures destabilizes most models; a linear-coupling subclass still gives inflation.","key_machinery":"The load-bearing object is the scalar-tensor localization of the non-local action: each $\\Box^{-1}$ is turned into an auxiliary scalar field constrained by a Lagrange multiplier, so the original action becomes a multi-field theory with kinetic matrices $K_1$ and $K_2$ in the Einstein frame. The decisive conditions are the Legendre inversion requirement $F_{\\chi\\chi}F_{\\eta\\eta}-F_{\\chi\\eta}^2\\neq0$, which distinguishes the generic ghost-ridden case from degenerate models, and Sylvester's criterion, the standard determinant conditions for a symmetric matrix to be positive definite, applied to $K_1$ and $K_2$, which converts ghost freedom into inequalities on the scalar fields and on $G'$. For the inflationary analysis, the central object is the Einstein-frame potential $Y_2(\\Phi,\\Xi,\\Psi)=\\frac{V_0\\chi_2^2(\\Psi)+\\frac{\\left(e^{\\sqrt{2/3}\\Phi}+\\Xi^2/6-a_2\\right)^2}{4b_2}}{e^{2\\sqrt{2/3}\\Phi}}$, whose plateau and minimum govern the slow-roll phase, together with the no-ghost inequality that confines the field-space trajectory.","core_discovery":"Working from the action $F(R,\\mathcal{R},\\Box^{-1}R,\\ldots,\\Box^{-m}R,\\Box^{-1}\\mathcal{R},\\ldots,\\Box^{-n}\\mathcal{R})$, the paper localizes each inverse d'Alembert operator with auxiliary scalars and Lagrange multipliers, and rewrites the theory in the Einstein frame. The central claim is that no matter the form of $F$, the kinetic sector of this scalar-tensor theory contains $N=m+n$ ghost fields whenever the Legendre inversion condition $F_{\\chi\\chi}F_{\\eta\\eta}-F_{\\chi\\eta}^{2}\\neq0$ holds; a purely Palatini truncation does not avoid them, since the Palatini scalar loses dynamics but the non-local sectors still contribute ghosts. The ghost-free escape is degeneracy: actions in which local and non-local parts are carried by different curvatures, namely $\\mathcal{L}_{1}=f(\\mathcal{R})+\\mathcal{R}G(\\Box^{-1}\\mathcal{R})-V(\\Box^{-1}\\mathcal{R})$ and $\\mathcal{L}_{2}=f(R)+R G(\\Box^{-1}R)-V(\\Box^{-1}R)$, violate that inversion condition and leave exactly three propagating scalars. Sylvester's criterion on the kinetic matrices yields the no-ghost inequalities $\\phi>0,\\ \\xi<0,\\ G'(\\alpha)>(\\phi-\\xi)/6$ for $\\mathcal{L}_1$ and $\\phi>0,\\ \\psi<0,\\ G'(\\beta)>-\\psi/6$ for $\\mathcal{L}_2$. For a flat FLRW background and quadratic $f(R)$, the $\\mathcal{L}_2$ model produces a Starobinsky-like plateau potential deformed by the non-local coupling; numerical integration shows the field $\\Phi$ drives inflation, $\\Xi$ settles to zero, and $\\Psi$ freezes as a light spectator, with the no-ghost condition $\\frac{1}{6}\\left(\\frac{d\\chi_2}{d\\Psi}\\right)^2\\left(\\sigma_2 e^{\\sqrt{2/3}\\Phi}+\\frac{\\Xi^2}{6}\\right)<1$ satisfied along the trajectory.","pith_inferences":["Beyond the paper, the light spectator $\\Psi$ in the $\\mathcal{L}_2$ model should generate isocurvature and non-Gaussian perturbations whose amplitude the paper does not compute; computing the primordial power spectrum would provide a sharper observational test than the background-level e-fold count.","The degeneracy that makes $\\mathcal{L}_1$ and $\\mathcal{L}_2$ ghost-free means the scalar-tensor representation is non-invertible at the Legendre-transformation level, so a direct analysis of the original non-local action could reveal whether ghost freedom survives the localization step.","The paper's classification suggests a constructive pattern: among hybrid non-local actions, stability selects those whose non-local part is linearly coupled to the curvature of opposite type to the local $f$; testing higher-order or multi-copy versions of $G(\\Box^{-1}R)$ would check whether this pattern persists.","Because the no-ghost window ties $\\Xi$ and $\\Psi$ together, the spectator could leave a measurable imprint on tensor-to-scalar ratio predictions; comparing the model's predictions with CMB bounds on $r$ and $n_s$ is a direct extension of the background analysis."],"forward_implications":["Every non-degenerate non-local hybrid action of the form (2.1) carries at least $m+n$ ghost fields, so this entire class is excluded as a stable gravitational theory unless extra mechanisms are introduced.","The degenerate Lagrangians $\\mathcal{L}_1$ and $\\mathcal{L}_2$ are ghost-free exactly when the Sylvester inequalities hold, giving a concrete recipe for building stable non-local hybrid models from known $f(R)$ actions.","Only the metric-$f(R)$ with Palatini non-localities (model $\\mathcal{L}_2$) yields a finite slow-roll phase; the Palatini-$f(\\mathcal{R})$ with metric non-localities (model $\\mathcal{L}_1$) has a potential flat in $\\Phi$, so it slow-rolls indefinitely and cannot reheat.","In $\\mathcal{L}_2$ with quadratic $f(R)$ and either power-law or exponential kinetic couplings, inflation is effectively single-field, driven by $\\Phi$, with $\\Xi$ at its minimum and $\\Psi$ acting as a light spectator damped by Hubble friction.","The non-local terms deform the Starobinsky-like potential and change the number of e-folds and the field trajectories, which in principle shifts the scalar spectral amplitude and other observables relative to Starobinsky inflation."],"supporting_citations":[{"why":"It supplies the purely metric non-local model whose localization procedure and ghost-counting result this paper extends, and it fixes the baseline that $N$ ghost fields appear.","marker":"[115]"},{"why":"It introduces hybrid metric-Palatini gravity, the framework whose additive hybridization scheme is used to build the ghost-free Lagrangians.","marker":"[98]"},{"why":"It provides the generalized hybrid metric-Palatini construction with the doubled scalar sector and the Einstein-frame reduction the paper adopts.","marker":"[99]"},{"why":"It gives the scalar-mode analysis of extended hybrid metric-Palatini gravity, including the sign constraints on the $\\Xi$ field that reappear in the ghost conditions.","marker":"[101]"},{"why":"It is the original non-local cosmology proposal based on $\\Box^{-1}R$, the ancestor of the non-local terms being localized here.","marker":"[110]"},{"why":"It presents an amended integral-kernel model and is cited as the route that keeps non-locality explicit rather than localizing it.","marker":"[114]"},{"why":"It defines the Starobinsky $R^2$ inflation model whose potential is the benchmark that the non-local terms deform.","marker":"[81]"},{"why":"It documents the Palatini $f(\\mathcal{R})$ formalism in which the curvature alone excites no additional scalar, the contrast case for the ghost counting.","marker":"[96]"}],"fun_headline_variants":["Ghosts die when local and non-local parts split in hybrid gravity","Non-local hybrid gravity: ghosts unless actions are degenerate","Inflation survives in degenerate non-local hybrid gravity","Hybrid gravity needs degeneracy to avoid ghosts and inflate","Special non-local gravity models evade ghosts and inflate"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is that replacing $\\Box^{-1}$ with auxiliary scalar fields gives an exact, boundary-condition-independent reformulation of the non-local action; if that localization is only formal, or if the choice of boundary conditions for the inverse d'Alembert operator changes the dynamics, the ghost count and no-ghost inequalities derived in the scalar-tensor picture do not apply to the original non-local theory.","fun_headline_variants_meta":{"raw":{"variants":["Ghosts die when local and non-local parts split in hybrid gravity","Non-local hybrid gravity: ghosts unless actions are degenerate","Inflation survives in degenerate non-local hybrid gravity","Hybrid gravity needs degeneracy to avoid ghosts and inflate","Special non-local gravity models evade ghosts and inflate"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000308,"raw_usage":{"total_tokens":1851,"prompt_tokens":1123,"completion_tokens":728,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":739,"completion_tokens_details":{"reasoning_tokens":648}},"tokens_in":739,"tokens_out":728,"duration_ms":6508,"temperature":1.0,"reasoning_tokens":648,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-11T11:39:32.335444+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Take the simplest non-degenerate case $m=n=1$ in action (2.1) and perform a Hamiltonian or Ostrogradski constraint analysis directly on the non-local action, without localizing $\\Box^{-1}$; if the ghost number is not $2$, or if it changes when boundary conditions of the inverse d'Alembert operator are varied, the stability claim fails. A second decisive test is to exhibit any non-degenerate hybrid action of the form (2.1) that is ghost-free, which would directly contradict the claimed unavoidability of $m+n$ ghosts.","supporting_citations":[{"cited_title":"Ghosts in classes of non-local gravity","cited_arxiv_id":"1412.1575","evidence_quote":"It supplies the purely metric non-local model whose localization procedure and ghost-counting result this paper extends, and it fixes the baseline that $N$ ghost fields appear."},{"cited_title":"Generalized hybrid metric-Palatini gravity","cited_arxiv_id":"1302.2355","evidence_quote":"It provides the generalized hybrid metric-Palatini construction with the doubled scalar sector and the Einstein-frame reduction the paper adopts."}],"review_version":1}