{"id":"a43ff69b-559a-4771-bb9c-f680311695e8","arxiv_id":"2610.02735","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"low","formal_verification":"none","parameter_count":4,"one_line_summary":"A quadratic non-Gaussian ansatz in GW helicity amplitudes fails polarization-frame covariance in PTA bispectra, while a detector-projected construction yields a timing residual skewness of order 10^-15 in Gaussian-rms units.","lead":"This theory paper studies three-point statistics of a gravitational wave background in pulsar timing arrays. It shows the standard helicity-space non-Gaussian ansatz is not polarization-frame covariant, and builds a detector-projected model whose timing residual skewness is tiny.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek/deepseek-v4-flash","headline":"The benchmark skewness (90), Fig. 1 rests on the per-direction channel ansatz (61), not shown—and likely impossible by the (44) phase argument—to be a projection of any single tensor-field nonlinearity; the nonzero µ3 may then be a property of imposed channel statistics, not of a physical SGWB.","rationale":"The paper is mathematically careful, internally consistent, and unusually explicit about its own limitations; the reader's CONDITIONAL verdict is well grounded. My pass agrees with the reader's weakest_assumption: the detector-projected channels (60)-(61) are not shown (b174) to be projections of a single physical tensor field, and the quantitative content of the positive result hangs on that condition. I sharpen it into a structural point: the mechanism that makes (61) polarization-frame invariant—detector projections inside each convolution factor, so gauge phases cancel factor-by-factor at internal directions—is the same feature that prevents (61) from being the projection of a common nonlinear tensor field, whose projection can only sit at the output direction k̂. The (44) phase analysis indicates no local quadratic convolution of helicity modes can be covariant, so the spin-covariant construction deferred to b293 is not a formality. The Δ-test cleanly distinguishes 'not yet shown' from 'cannot', and the projected-(34) recomputation gives a decisive contrast against Fig. 1. I found no internal inconsistency: the angular identities (77), (83)-(84) and the Υ reduction (86)-(88) check out; the terse sign-sector reduction to (90) is plausible and, if subtly wrong, would alter the benchmark by O(1) factors only, not its order of magnitude; the magnitude chain µ3 ~ 10^-36 s^3, K ~ 10^-15 against σ(r³) ~ 4×10^-21 s^3 is consistent. The negative result (Section III B) is independent of all benchmark approximations. Because the paper explicitly scopes the construction as a nonstationary, detector-level, phenomenological template (b176, b183, b304), its claims as written survive, and the verdict remains CONDITIONAL; the test above is the natural gate for the promised tensor-field formulation (b293), not a ground for rejection.","tokens_in":35897,"tokens_out":24333,"duration_ms":242079,"concrete_test":"Check Δ: compute ⟨|Δ_A(k,ê)|²⟩ and ⟨Δ_A(k,ê₁)Δ*_A(k,ê₂)⟩ over the Gaussian seed, with Δ_A as defined in the attack (using eqs. 59, 61, 34 with the band/PSD of §IV). If the variance is nonzero, (61) is not the projection of the scalar-convolution tensor ansatz; combined with the (44) phase argument this resolves the realizability question deferred at b174/b293. Then recompute the Fig. 1 third moment with the alternative channel D_A(k̂,ê)·g_NL ∫dp h^L_A(k+p) h^{L*}_A(p) (projection of (34) at the output direction), same A_gw, g_NL, band, epochs: if it is polarization-basis dependent or vanishes while (90) gives ~10^-36 s^3, the nonzero skewness is a property of (61) alone, not of any quadratic tensor-field SGWB model.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The positive claim—nonzero, polarization-frame-independent µ3 ~ 10^-36 s^3, |K| ~ 10^-15 (Fig. 1; eqs. 90, 100-102)—is a property of the channel ansatz (60)-(61), imposed separately for each pulsar direction ê. The authors concede (b174) they have not shown these channels are projections of a single detector-independent nonlinear tensor field. The concern is sharper than a missing proof. Eq. (44) shows a scalar-style helicity convolution is not polarization covariant: its integrand carries phase exp(-is_A[α(k̂+p̂)-α(p̂)]) instead of the required exp(-is_A α(k̂)). Eq. (61) achieves covariance by placing the detector projection inside each factor, so D_A((k+p)̂,ê)D_Ā(p̂,ê) cancels both phases factor-by-factor; a projection of any common tensor nonlinearity h^NL_A(k) = g_NL ∫ dp h^L_A(k+p) h^{L*}_A(p) would carry its projection D_A(k̂,ê) at the output direction only. The two constructions differ by Δ_A(k,ê) = ∫dp h^L_A(k+p) h^{L*}_A(p)[D_A((k+p)̂,ê)D_Ā(p̂,ê) - D_A(k̂,ê)] plus unequal subtraction terms, whose Gaussian variance is generically nonzero; and the (44) phase argument leaves no room for a local quadratic kernel to repair covariance. Thus the nonzero benchmark is most plausibly a channel-level property with no tensor-field SGWB realization, consistent with the vanishing result of ref. [97] under its assumptions. This does not touch the negative result (Section III B), which is structural and clean, nor the internal algebra: I spot-checked (77), (83)-(84), (86), (91), and the magnitude chain ⟨r²⟩ ~ 10^-14 s² → σ(r³) ~ 4×10^-21 s³ → µ3 = Kσ ~ 4×10^-36 s³, all mutually consistent. The paper acknowledges the gap (b174-b176, b183, b293), so this is a scoping concern; it remains the load-bearing condition of the central quantitative claim.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek/deepseek-v4-flash","summary":"The paper studies perturbative departures from Gaussian statistical isotropy in the stochastic gravitational wave background as probed by pulsar timing arrays, using a Gaussian seed plus a quadratic non-Gaussian correction. It reports two main results. First (Section III B), the helicity-space quadratic ansatz (33)-(34) produces a timing-residual bispectrum that is not manifestly invariant under local polarization-basis rotations for tensor and vector modes: the response coefficients J^A_lm carry spin phases (eqs. 56-57) that the helicity bispectrum (41) does not compensate, with the obstruction made explicit by the phase transformation (44). Second (Section III C), the paper constructs a detector-projected channel ansatz (60)-(61) in which the response factor D_A is placed inside each quadratic factor, so spin phases cancel factor-by-factor; within this construction, and with an Earth-term-only, simplified-response, isotropic, unpolarized benchmark, the single-pulsar timing-residual third moment is nonzero and polarization-frame independent (eq. 90). Numerical benchmarks (Fig. 1) give µ3 ~ 10^-36 s^3 and |K| ~ 10^-15 for A_gw = 10^-15, g_NL = 1.","tokens_in":36288,"tokens_out":12742,"duration_ms":81839,"significance":"If the negative result stands, it is a genuine and useful cautionary result: it demonstrates that the naive scalar-style quadratic convolution used in some of the PTA non-Gaussianity literature is inconsistent for three-point timing-residual observables, and it identifies the required spin-covariant structure. That result is structural, clean, and checkable, and is appropriately distinguished from the vanishing-detector-correlation result of ref. [97]. The positive construction is internally consistent: the angular integrals (77)-(86), the closed-form kernel (87) with the quartic form (C3), and the reduction to (90)-(91) are explicit and computable, and I found no error in the algebra I checked. The scaling K ∝ g_NL A_gw (102) is largely a dimensional consequence of the ansatz, and the numerical benchmark therefore determines only coefficients and time dependence. The open question is whether the detector-projected channel ansatz (61) has any realization as a projection of a single nonlinear tensor field; the authors acknowledge they have not shown this (b174-b175), and the phase structure of (44) suggests a local quadratic kernel cannot repair the covariance.","major_comments":[{"comment":"The physical status of the positive benchmark needs to be resolved or the paper reframed. The channel ansatz (61) achieves polarization covariance by placing the response D_A(k̂,ê) inside each quadratic factor, so that the spin phases cancel factor-by-factor via D_A((k+p)̂,ê)D_Ā(p̂,ê). A projection of any tensor-field nonlinearity of the form h^NL_A(k) = g_NL ∫ dp h^L_A(k+p) h^{L*}_A(p) would carry the projection D_A(k̂,ê) only at the output direction. The two integrands differ by Δ_A(k̂,ê) = ∫ dp h^L_A(k+p)h^{L*}_A(p)[D_A((k+p)̂,ê)D_Ā(p̂,ê) − D_A(k̂,ê)] plus unequal subtraction terms, whose Gaussian variance is generically nonzero, and the phase argument of (44) leaves no room for a local quadratic kernel to repair the covariance. The authors disclose the missing realization (b174-b175), but the abstract and the Conclusions (b289, b298) still list the nonzero skewness as a main result. The revision should either (i) prove or disprove that the channels in (61) can be written as projections of a single nonlinear tensor field and, if they cannot, state explicitly that the nonzero µ3 is a property of the imposed channel statistics rather than of any SGWB - consistent with the vanishing result of ref. [97] - or (ii) restructure the paper so the polarization-covariance obstruction of Section III B is the headline result and the detector-projected benchmark is presented explicitly as a formal consistency check of a covariant channel-level construction.","section":"Section III C and Conclusions (eqs. 60-61, 70; b174-b175; b289)","authors_acknowledge":"Once (61) is imposed separately for each direction ˆe, the nonlinear channels are a phenomenological model; we have not shown that they can all be written as projections of a single detector-independent nonlinear tensor field.","authors_acknowledge_p":0.85},{"comment":"The headline numerical values are quoted without the qualifiers that are disclosed elsewhere in the paper. The Conclusions state that 'for the existing unit-coupling benchmark with A_gw = 10^-15, the third moment is of order 10^-36 s^3 and its Gaussian-rms-normalized magnitude reaches order 10^-15' (b299). That number rests on four separate restrictions: the channel-level ansatz (61) whose tensor-field realizability is open (b174), the equal-time nonstationary construction that drops the signed-frequency stationarity condition (b177, b183), the Earth-term-only reduction and simplified response D_A/2 whose cubic-statistic corrections are uncalculated and not assumed order-one (b196-b198), and a specific frequency prescription (two power-spectrum legs in [f_min, f_max], the third over the triangle-supported range). Each of these is disclosed in Sections III C and IV, but the Conclusions should recapitulate them in the sentence that quotes the numbers, so that a reader taking the abstract or the conclusions alone does not mistake the benchmark for a prediction for an isotropic tensor SGWB.","section":"Conclusions (b299) and Section IV B","authors_acknowledge":null,"authors_acknowledge_p":null}],"minor_comments":[{"comment":"The repeated use of 'existing' ('the existing numerical benchmark', 'the existing evaluation of (90)', 'the existing frequency prescription', 'the retained curves') reads as leftover from an iterative drafting process; recommend neutral phrasing such as 'our benchmark' and 'the evaluation of (90)'.","section":"Section IV headings and text","authors_acknowledge":null,"authors_acknowledge_p":null},{"comment":"The quantity K = γ1/√15 is a purely conventional rescaling of the dimensionless skewness by the inverse square root of the Isserlis factor appearing in Var(r³); the text should state plainly that K is not an estimator signal-to-noise ratio, which I believe is the intent but is worth making explicit at the definition.","section":"Eq. (100)","authors_acknowledge":null,"authors_acknowledge_p":null},{"comment":"The statement that determining the compensating spin factors 'is left for future work' could be sharpened: the obstruction at (44) is that the phase mismatch is nonlocal in direction space, α((k+p)̂) − α(p̂) ≠ α(k̂) generally, so the required kernel is necessarily spin-weighted and direction-sensitive; one sentence identifying this nonlocality would make the difficulty concrete.","section":"Section III B (b156-b157)","authors_acknowledge":null,"authors_acknowledge_p":null},{"comment":"The caption should state explicitly that only γ_gw varies among the four curves and that A_gw = 10^-15 and g_NL = 1 are fixed in all curves, so the reader does not need to infer it from the text.","section":"Figure 1 caption","authors_acknowledge":null,"authors_acknowledge_p":null},{"comment":"The three-point result (41), which seeds all subsequent Wick contractions, is introduced with 'it can be shown'; a one-line derivation showing the two contractions leading to the two terms P_{A Ā''}P_{A' A''} and P_{A A''}P_{A' Ā''} would improve readability, although the result itself checks out.","section":"Eq. (41)","authors_acknowledge":null,"authors_acknowledge_p":null}],"recommendation":"major_revision","confidential_remarks":"The paper is technically careful, the negative result is clean and publishable, and the authors are unusually candid about the limitations of the positive benchmark. My main concern is structural: the nonzero skewness is presented in the abstract and conclusions as a headline result, but the paper's own disclosure (b174) plus the phase argument of (44) make it likely that the detector-projected channel ansatz (61) has no realization as a projection of a single nonlinear tensor field. If that is the case, the benchmark is a property of the imposed channel statistics, and the paper should be reframed around the Section III B result, with the benchmark clearly demoted. I would ask the editor to ensure the revision explicitly resolves this - either by a realizability argument or by the reframing - rather than leaving it as an acknowledged-but-unaddressed gap. The 'existing' language in Section IV also suggests the benchmarks were retrofitted during drafting; a final polish pass is needed."},"author_rebuttal":null,"desk_editor":{"model":"deepseek/deepseek-v4-flash","letter":"Two things to know. First, the negative result is real: Section III B shows the helicity-space quadratic ansatz (33-34) fails polarization-frame covariance in the PTA timing-residual bispectrum. This is structural and clean. The response coefficients (56-57) carry spin phases that the helicity bispectrum in (41) does not compensate, and I checked the logic carefully; it holds up. That alone is a useful result worth having on record, because it rules out the naive scalar-style convolution for PTA three-point observables.\n\nSecond, the constructive benchmark in Section III C is well-hedged, internally consistent, and physically unverified. The detector-projected channels (61) cancel the spin phases factor-by-factor, and the angular algebra checks out: I spot-checked (77), (83-84), (86), (91), and the magnitude chain linking <r2> ~ 10^-14 s2 to mu3 ~ 4 x 10^-36 s3. The benchmark yields |K| ~ 10^-15 for unit coupling, and the authors signal its status honestly throughout with 'approximately equals' and explicit enumeration of every approximation.\n\nThe load-bearing soft spot is the physical status of the channels in (61), and it is a real one. The construction is imposed separately for each pulsar direction, and the authors concede (b174) they have not shown these channels are projections of a single detector-independent nonlinear tensor field. The stress-test sharpens this beyond a missing proof: a projection of any common tensor-field nonlinearity would carry the response D_A(k, e) only at the output direction, while (61) places it inside each factor. The two constructions differ by a term with generically nonzero Gaussian variance, and the (44) phase argument leaves no room for a local quadratic kernel to repair the covariance. So the nonzero third moment is most plausibly a property of the imposed channel statistics rather than of a physical SGWB. The authors are explicit about the gap (b174-b176, b183, b293), which makes this a scoping concern rather than a missed flaw, but it is exactly the condition the quantitative claim rests on.\n\nTwo smaller points, both also disclosed by the authors: the benchmark keeps only the Earth term and replaces the antenna response with a simplified projection (b196-b198), and the cubic-statistic corrections are uncalculated and not assumed order-one. Neither touches the negative result. The citation pattern is fine; the nonvanishing claim is explicitly set against the vanishing result of Jimenez Cruz et al [97] with the differences stated.\n\nFor a PTA theory group working on SGWB statistics, this is worth a reading. It deserves a serious referee: the negative result alone justifies review, and the benchmark is a useful template even with its physical status unresolved. The referee's job is to push on the connection to a physical tensor field, which the authors defer to future work.","headline":"The polarization-frame covariance critique is clean and structural; the nonzero skewness benchmark is an explicitly acknowledged channel-level construction whose connection to a physical SGWB remains open.","tokens_in":36888,"tokens_out":2378,"would_cite":true,"duration_ms":16668,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek/deepseek-v4-flash","headline":"The paper shows that the simple CMB-inspired quadratic non-Gaussian ansatz for gravitational-wave amplitudes fails a polarization-frame test, and that rebuilding it at detector level gives a nonzero but extremely small timing-residual…","keywords":["pulsar timing arrays","stochastic gravitational wave background","non-Gaussianity","timing residual skewness","polarization basis invariance","detector-projected amplitudes","bispectrum","primordial non-Gaussianity"],"falsifier":"Compute the one-pulsar timing-residual third moment (90) with the full antenna response and pulsar terms instead of the simplified $d_{ij}e^A_{ij}/2$ and Earth-term-only prescription; if the nonzero skewness vanishes or its size changes dramatically, the benchmark is an artifact. A measurement of $|K|$ with uncertainty below roughly $10^{-15}$ that finds no third moment would also contradict the unit-coupling $A_{\\rm gw}=10^{-15}$ benchmark.","tokens_in":37180,"feed_emoji":"📡","tokens_out":7147,"duration_ms":42613,"temperature":0.7,"pith_summary":"The paper asks whether a CMB-style quadratic non-Gaussian correction to a Gaussian gravitational-wave background can produce a detectable three-point skewness in pulsar-timing residuals. Its first result is negative: when the quadratic ansatz is written in the gravitational-wave helicity amplitudes, the resulting timing-residual bispectrum is not invariant under local rotations of the polarization basis for tensor and vector modes, so the simple scalar-inspired ansatz is inconsistent. Its second result is constructive: writing the same quadratic structure in the detector-projected amplitudes makes the polarization phases cancel, and the one-pulsar timing-residual third moment is nonzero, with unit coupling giving a third moment of order $10^{-36}\\,\\mathrm{s}^3$ and a Gaussian-rms-normalized magnitude $|K|$ of order $10^{-15}$ for $A_{\\rm gw}=10^{-15}$. The normalized skewness scales as $K\\propto g_{\\rm NL}A_{\\rm gw}$, so the small strain amplitude makes a small skewness natural in this perturbative model.","feed_headline":"Quadratic GW non-Gaussianity fails a frame test; skewness is tiny","feed_subtitle":"A CMB-style ansatz is not polarization-invariant for pulsar timing; a detector-level rebuild gives |K| ~ 10^-15.","key_machinery":"The central object is the detector-projected gravitational-wave amplitude $\\tilde{h}_A(\\vec{k},\\hat{e}) = h_A(\\vec{k})\\,d_{ij}(\\hat{e})\\,e^A_{ij}(\\hat{k})$, where $d_{ij}(\\hat{e}) = \\hat{e}_i\\hat{e}_j$ is the pulsar detector tensor. Under a local rotation of the polarization basis, the helicity amplitude and the polarization tensor acquire opposite spin-dependent phases, so their product is invariant; imposing the quadratic non-Gaussian ansatz on this projected amplitude cancels the phases explicitly in the three-point correlation. The quantitative evaluation uses the simplified response $F^A \\sim d_{ij} e^A_{ij}/2$, retains only the Earth term, and reduces the angular integrals to the analytic kernel $\\Upsilon(f,f',f'')$ built from $J_{\\ell\\ell 0}(f,f',f'')$ with triangle support $β(\\Delta)$. This machinery is what turns the polarization-frame problem into a concrete, evaluable single-pulsar skewness.","core_discovery":"The central claim is twofold. First, the helicity-space ansatz (33–34), in which the nonlinear part is a quadratic convolution of Gaussian helicity amplitudes, does not produce a manifestly polarization-frame-invariant timing-residual bispectrum for tensor and vector modes: under a local polarization-basis rotation the quadratic convolution lacks the spin-dependent compensating phase required by the response coefficients, so the bispectrum (53) depends on the basis choice. Second, defining the same quadratic structure on the detector-projected amplitude $\\tilde{h}_A(\\vec{k},\\hat{e}) = h_A(\\vec{k})\\, d_{ij}(\\hat{e})\\, e^A_{ij}(\\hat{k})$ makes the phase cancellation explicit, and the resulting one-pulsar timing-residual third moment is nonzero and polarization-frame independent. With an isotropic, unpolarized tensor background, equal helicity powers, and the paper's simplified Earth-term response, the benchmark gives a third moment of order $10^{-36}\\,\\mathrm{s}^3$ and a Gaussian-rms-normalized magnitude $|K|$ of order $10^{-15}$ for $A_{\\rm gw}=10^{-15}$ and $g_{\\rm NL}=1$, with $K\\propto g_{\\rm NL}A_{\\rm gw}$. The paper stresses that this is an approximate, detector-channel-level phenomenological benchmark, not a generic prediction for an isotropic tensor stochastic background.","pith_inferences":["Editorial inference: Restoring the full antenna response and pulsar terms may change the third moment by more than an order of magnitude, so the reported $10^{-36}\\,\\mathrm{s}^3$ and $|K|\\sim10^{-15}$ should be read as an order-of-magnitude benchmark rather than a sharp prediction.","Editorial inference: The frame-invariance failure likely attaches to any scalar-inspired quadratic convolution of helicity amplitudes, not just the one written here, so spin-covariant structure is a general requirement for tensor and vector non-Gaussian SGWB models.","Editorial inference: A single-pulsar skewness this small will be hard to detect; higher connected moments such as the four-point function, or multi-pulsar bispectrum combinations, may be better observational targets even when the bispectrum is zero.","Editorial inference: The detector-projected construction could be adapted to vector and scalar polarizations in alternative theories of gravity, where the same polarization-phase issue appears."],"forward_implications":["The simple quadratic helicity ansatz (33–34) is not usable as written for PTA three-point statistics; consistent tensor and vector non-Gaussian models must carry compensating spin phases.","Non-Gaussian timing-residual statistics can be formulated at detector level, where polarization-frame invariance is automatic, and that formulation predicts a nonzero single-pulsar skewness.","At unit coupling and $A_{\\rm gw}=10^{-15}$, the third moment is about $10^{-36}\\,\\mathrm{s}^3$ and $|K|\\sim10^{-15}$, scaling as $K\\propto g_{\\rm NL}A_{\\rm gw}$.","Because the two-point function is unchanged at $\\mathcal{O}(g_{\\rm NL})$, a measured third moment would probe phase-coupled non-Gaussian information absent from the power spectrum.","Within the illustrated band, redder spectral slopes increase the magnitude of the third moment."],"supporting_citations":[{"why":"Supplies the CMB quadratic bispectrum template that motivates and defines the non-Gaussian ansatz transplanted to gravitational-wave amplitudes.","marker":"[85]"},{"why":"Extends the CMB bispectrum formalism and is used as the template for measuring primordial non-Gaussianity.","marker":"[86]"},{"why":"The stationary three-point result whose assumptions differ; the paper contrasts its nonstationary detector-level benchmark against this work.","marker":"[97]"},{"why":"Isserlis' theorem is used to evaluate Gaussian higher moments and to normalize the skewness through $\\langle r^6\\rangle = 15\\langle r^2\\rangle^3$.","marker":"[74]"},{"why":"Supplies the generalized projection factors used in the harmonic-space timing-residual and bispectrum expressions.","marker":"[101]"},{"why":"Provides the subluminal stochastic gravitational-wave response coefficients and the spin-phase structure behind the polarization-frame calculation.","marker":"[102]"},{"why":"States the local polarization-basis rotation law for helicity amplitudes and polarization tensors used to identify the missing compensating phases.","marker":"[103]"},{"why":"Constructs filters that select momentum-triangle configurations, the class of configurations over which the equal-time benchmark integrates.","marker":"[95]"},{"why":"Studies stationary non-Gaussian backgrounds with folded configurations; the paper explicitly does not impose the signed-frequency stationarity condition assumed there.","marker":"[96]"}],"fun_headline_variants":["Pulsar timing skewness from GW non-Gaussianity is tiny","Quadratic GW non-Gaussianity fails polarization frame test","Pulsar timing third moment: frame-invariant but minuscule","Non-Gaussian GW ansatz not invariant; skewness near 1e-15","Weak non-Gaussianity in pulsar arrays yields negligible skew"],"cache_read_input_tokens":0,"weakest_assumption_plain":"The nonzero benchmark skewness stands on treating the detector-projected nonlinear channels (61) as a physical description of non-Gaussian gravitational-wave statistics, which the paper explicitly says it has not shown; the magnitude also rests on the Earth-term-only, simplified-antenna-response approximations, whose corrections to the cubic statistic are not calculated.","fun_headline_variants_meta":{"raw":{"variants":["Pulsar timing skewness from GW non-Gaussianity is tiny","Quadratic GW non-Gaussianity fails polarization frame test","Pulsar timing third moment: frame-invariant but minuscule","Non-Gaussian GW ansatz not invariant; skewness near 1e-15","Weak non-Gaussianity in pulsar arrays yields negligible skew"]},"model":"deepseek/deepseek-v4-flash","effort":"low","cost_usd":0.000148,"raw_usage":{"cost":0.000147868,"is_byok":false,"cost_details":{"upstream_inference_cost":0.000147868,"upstream_inference_prompt_cost":0.000106708,"upstream_inference_completions_cost":4.116e-05},"total_tokens":1114,"prompt_tokens":967,"completion_tokens":147,"prompt_tokens_details":{"audio_tokens":0,"video_tokens":0,"cached_tokens":256,"cache_write_tokens":0},"completion_tokens_details":{"audio_tokens":0,"image_tokens":0,"reasoning_tokens":53}},"tokens_in":967,"tokens_out":147,"cost_source":"openrouter_usage","duration_ms":2257,"image_pages":[],"temperature":1.0,"billed_attempts":1,"reasoning_tokens":53,"transport_provider":"openrouter","cache_read_input_tokens":0,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-10-05T19:04:13.776314+00:00","model_set":{"reader":"deepseek/deepseek-v4-flash"},"falsifier":"Compute the one-pulsar timing-residual third moment (90) with the full antenna response and pulsar terms instead of the simplified $d_{ij}e^A_{ij}/2$ and Earth-term-only prescription; if the nonzero skewness vanishes or its size changes dramatically, the benchmark is an artifact. A measurement of $|K|$ with uncertainty below roughly $10^{-15}$ that finds no third moment would also contradict the unit-coupling $A_{\\rm gw}=10^{-15}$ benchmark.","supporting_citations":[{"cited_title":"Acoustic Signatures in the Primary Microwave Background Bispectrum","cited_arxiv_id":"astro-ph/0005036","evidence_quote":"Supplies the CMB quadratic bispectrum template that motivates and defines the non-Gaussian ansatz transplanted to gravitational-wave amplitudes."},{"cited_title":"Measuring primordial non-Gaussianity in the cosmic microwave background","cited_arxiv_id":"astro-ph/0305189","evidence_quote":"Extends the CMB bispectrum formalism and is used as the template for measuring primordial non-Gaussianity."},{"cited_title":"New test of modified gravity with gravitational wave experiments","cited_arxiv_id":"2509.08273","evidence_quote":"The stationary three-point result whose assumptions differ; the paper contrasts its nonstationary detector-level benchmark against this work."},{"cited_title":"Pulsar-timing arrays, astrometry, and gravitational waves","cited_arxiv_id":"1810.02369","evidence_quote":"Supplies the generalized projection factors used in the harmonic-space timing-residual and bispectrum expressions."},{"cited_title":"Subluminal stochastic gravitational waves in pulsar-timing arrays and astrometry","cited_arxiv_id":"2007.11009","evidence_quote":"Provides the subluminal stochastic gravitational-wave response coefficients and the spin-phase structure behind the polarization-frame calculation."},{"cited_title":"Linear polarization of the stochastic gravitational-wave background with pulsar timing arrays","cited_arxiv_id":"2312.03056","evidence_quote":"States the local polarization-basis rotation law for helicity amplitudes and polarization tensors used to identify the missing compensating phases."},{"cited_title":"Searching for Bispectrum of Stochastic Gravitational Waves with Pulsar Timing Arrays","cited_arxiv_id":"1812.10615","evidence_quote":"Constructs filters that select momentum-triangle configurations, the class of configurations over which the equal-time benchmark integrates."},{"cited_title":"Probing a stationary non-Gaussian background of stochastic gravitational waves with pulsar timing arrays","cited_arxiv_id":"1910.04758","evidence_quote":"Studies stationary non-Gaussian backgrounds with folded configurations; the paper explicitly does not impose the signed-frequency stationarity condition assumed there."}],"review_version":1}