{"id":"9e8c13b0-5c03-49f3-ad4a-55ae2e042ffb","arxiv_id":"1403.4596","paper_version":4,"verdict":"CONDITIONAL","confidence":"HIGH","novelty_score":4.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"A step-like drop in the inflaton potential's slope, suppressing large-scale scalar power by about 26%, fits Planck plus BICEP2 data better than a running spectral index by a change in chi-squared of about 7.5 for one extra parameter.","lead":"The authors test whether a brief faster-roll phase in the early universe, which suppresses primordial lumpiness at the largest observable scales, can reconcile the BICEP2 tensor signal with the Planck temperature limits. They find it fits the combined data better than the standard running-of-the-spectral-index explanation, at the cost of one extra parameter.","discovery_kind":"unclear","skeptic_critique":{"model":"selfhosted/deepseek-v4.1-flash","headline":"Table 2's relative-likelihood column evaluates exp(ΔAIC) rather than the exp(ΔAIC/2) specified in Eq. 2.5, overstating the AIC preference for the suppression model by roughly a factor of 20–200; the fit direction survives, but the 'strongly favoured' claim is quantitatively inflated.","rationale":"The reader's verdict of CONDITIONAL is appropriate, and the reader's fixed-parameter concern is legitimate: Section 2 explicitly holds all cosmological and nuisance parameters other than τ, n_s, ln(10¹⁰A), r, k*, and Δ fixed at Planck 'base r' best-fit values, and no full-parameter run is reported. However, that concern is probably subdominant for the central claim because the suppression model's improvement is localized at ℓ < 30, where ω_b, ω_c, and θ are weakly constrained by the data and the Sachs-Wolfe plateau depends mainly on A_s, n_s, and τ, all of which are varied. The more concrete and checkable issue the reader missed is the factor-of-two inconsistency between Eq. 2.5 and Table 2: the tabulated relative likelihoods are exp(ΔAIC), not exp(ΔAIC/2), which inflates the AIC preference for suppression by roughly 20–200. This is a definite arithmetic error, not a conceptual one, and it does not reverse the direction of the model comparison; the suppression model remains favoured, just less dramatically. The paper's second result, that a primordial scalar-tensor correlator drops out of the TT power spectrum (Eqs. 3.4–3.5), is a clean derivation and the authors are explicit that it reverses their earlier conclusion after refs. 33 and 34. The premise of the whole exercise, a primordial r near 0.2, was already under pressure from dust at the time of the final version, and the paper acknowledges this in the Introduction. Overall, the central claim holds up qualitatively but the quantitative strength is overstated; the verdict should remain CONDITIONAL, with the AIC correction as a specific revision item.","tokens_in":14723,"tokens_out":16218,"duration_ms":318390,"concrete_test":"Recompute Table 2's relative-likelihood column from the Table 1 χ² and Np values using Eq. 2.5. If the corrected entries are exp(ΔAIC/2) ≈ 0.006 and 0.062 rather than 2.45×10⁻⁵ and 0.0026, the 'strongly favoured' language in Section 2 and the Conclusions should be softened, and the abstract's 'considerably better' should be quantified as a likelihood ratio rather than left qualitative.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The paper's headline evidence for the suppression model is the AIC comparison in Table 2, which is meant to support the Δχ² = 7.5 improvement over the running model (Table 1). Eq. 2.5 defines the relative likelihood as L_AIC = exp[(AIC_min − AIC)/2]. Table 2 lists ΔAIC = −10.617 (ΛCDM+tensor) and −5.939 (running) with relative likelihoods 2.45×10⁻⁵ and 0.0026. These entries equal exp(ΔAIC), not exp(ΔAIC/2): ln(2.45×10⁻⁵) = −10.617 and ln(0.0026) = −5.95. Recomputing from the Table 1 χ² and Np values (AIC = 2Np + χ²) gives AIC_min − AIC = −10.234 and −5.556, so the correct relative likelihoods per Eq. 2.5 are exp(−5.117) ≈ 0.006 and exp(−2.778) ≈ 0.062. The suppression model is therefore favoured over running by roughly 16:1, not 385:1, and over ΛCDM+tensor by roughly 167:1, not 4×10⁴:1. The qualitative conclusion (suppression fits better) survives, but the 'strongly favoured' language and the abstract's 'considerably better' are overstated. Separately, the reader's fixed-parameter concern is acknowledged in Section 2 ('for simplicity') and is probably subdominant because the improvement is concentrated at ℓ < 30, where ω_b, ω_c, and θ matter little; but the paper does not demonstrate this, and the concentration of the gain at ℓ = 20–30 (cosmic-variance-limited, known-anomaly region) is not quantified.","agreement_with_reader":"partial"},"referee_report":{"model":"selfhosted/deepseek-v4.1-flash","summary":"The manuscript addresses the tension between the BICEP2 B-mode amplitude (r ≈ 0.2) and the Planck temperature-anisotropy limit (r < 0.11 under a constant tilt). Rather than invoking a large negative running, the authors study a Starobinsky-type step in the inflaton potential (Eq. 2.1) that suppresses scalar power by about 26% below k* ≈ 1.5×10⁻³ Mpc⁻¹, with best-fit Δ = 0.14 and r = 0.18. Fitting a reduced parameter space (τ, n_s, ln(10¹⁰A), r, k*, Δ) to Planck+WP+BICEP2, they report an effective χ² of 9838.87 versus 9846.43 for constant running and 9853.10 for ΛCDM+tensor (Table 1), and argue via AIC (Table 2) that the suppression model is strongly favoured. A second section argues that a primordial scalar-tensor anticorrelation, while it generates off-diagonal C_{ℓmℓ'm'}, leaves the temperature power spectrum C_ℓ unchanged and therefore cannot reconcile a large r with the TT limit. An appendix repeats the analysis with a step-function parametrisation and finds consistent parameters.","tokens_in":15043,"tokens_out":4821,"duration_ms":85669,"significance":"The question is timely and the proposed mechanism is physically motivated and simple: a modest, slow-roll-preserving change in the potential slope, rather than a running far outside slow-roll expectations. The negative result of Section 3 is useful and is presented with appropriate credit to Refs. [33,34] for correcting an earlier version (footnote 1), which is good practice and worth noting explicitly. The paper is clearly written, the model is falsifiable (the predicted low-ℓ suppression and the ringing in P_s are testable with Planck polarization and future B-mode data), and the parameter constraints in Figure 4 are reported with marginalized errors. The main strengths are the physical transparency of the suppression model and the honest correction of the earlier anticorrelation claim; the main weakness is that the headline statistical comparison rests on a reduced-parameter MCMC and on an AIC table whose likelihood column appears to be miscomputed.","major_comments":[{"comment":"The relative-likelihood column of Table 2 is inconsistent with Eq. (2.5). Eq. (2.5) defines L_AIC = exp[(AIC_min − AIC)/2], but the tabulated values equal exp(ΔAIC): ln(2.45×10⁻⁵) = −10.617 and ln(0.0026) = −5.95, matching the ΔAIC entries without the factor of 1/2. Recomputing from Table 1 (AIC = 2N_p + χ²) gives AIC_min − AIC = 10.234 for ΛCDM+tensor and 5.556 for running, so the correct relative likelihoods are approximately 0.006 and 0.062, not 2.45×10⁻⁵ and 0.0026. The suppression model is still preferred, but by roughly 16:1 over running rather than ~385:1. In addition, the tabulated ΔAIC values (−10.617, −5.939) do not reproduce the Table 1 χ² and N_p entries (which give −10.234 and −5.556), so the column needs to be recomputed from a single consistent set of numbers. The qualitative conclusion survives, but the phrases 'strongly favoured' (Section 2 and Conclusions) and 'considerably better' (abstract:s6) should be softened to match the corrected numbers.","section":"Table 2 and Eq. (2.5)","authors_acknowledge":null,"authors_acknowledge_p":null},{"comment":"The MCMC holds all cosmological and nuisance parameters other than τ, n_s, ln(10¹⁰A), r, k*, and Δ fixed at the Planck 'base r' best-fit values, as stated explicitly ('for simplicity'). The headline Δχ² ≈ 7.5 is small enough that shifts in ω_b, ω_c, θ, or the nuisance parameters could move it appreciably, and the paper does not demonstrate that the shortcut is harmless. Because the improvement is concentrated at 20 < ℓ < 30 (Figure 1), where the background parameters matter little, the concern is probably subdominant, but this should be shown rather than asserted. Either report a run with the full parameter space varied, or quantify how much of the Δχ² comes from the 20 < ℓ < 30 range and argue explicitly that the fixed parameters cannot absorb it. A short table decomposing the χ² improvement by multipole band would settle the point.","section":"Section 2, MCMC setup","authors_acknowledge":null,"authors_acknowledge_p":null}],"minor_comments":[{"comment":"The phrase 'a sudden change of the speed of the inflaton (by about 14%)' is imprecise: Δ = 0.14 is the fractional change in the slope of the potential (C₊ − C₋)/C₊, not the speed. The speed change is second order in Δ. Rephrase to 'a change in the slope of the inflaton potential' to match Eq. (2.1) and the text of Section 2.","section":"Abstract (abstract:s6)","authors_acknowledge":null,"authors_acknowledge_p":null},{"comment":"The caption states that 'Only the case with suppressed scalar power can accommodate the observed amplitude of r ~ 0.2', but the r column lists r = 0.18 for all three models, including ΛCDM+tensor and running. Clarify that the point is about fit quality at r = 0.18, not about which models can accommodate that value.","section":"Table 1 caption","authors_acknowledge":null,"authors_acknowledge_p":null},{"comment":"The marginalized constraint Δ = 0.106⁺⁰·⁰⁹²₋₀.₀₁₃ sits noticeably below the best-fit Δ = 0.14, and the errors are strongly asymmetric. Explain the shift (presumably a volume effect in the posterior) and state whether the best-fit or the marginalized value should be quoted as the headline suppression amplitude.","section":"Section 2, marginalized constraints","authors_acknowledge":null,"authors_acknowledge_p":null},{"comment":"The step-function fit gives k* = (4.07⁺⁰·⁷⁰₋₂.₄₅)×10⁻³ Mpc⁻¹, versus (1.80⁺⁰·⁸⁶₋₀.₆₉)×10⁻³ Mpc⁻¹ for the Starobinsky model. Calling these 'consistent' is a stretch given that the central values differ by a factor of ~2.3; state the overlap of the intervals explicitly rather than asserting consistency.","section":"Appendix A","authors_acknowledge":null,"authors_acknowledge_p":null},{"comment":"The negative result is derived under the Sachs-Wolfe approximation and a scale-invariant scalar-tensor correlator of the form (3.3). State these assumptions in the conclusion of the section, and note that the off-diagonal C_{ℓmℓ'm'} of Eqs. (3.4)–(3.5) are in principle observable even though C_ℓ is not affected; this would make the negative result more useful.","section":"Section 3","authors_acknowledge":null,"authors_acknowledge_p":null},{"comment":"The ℓ labels near the x-axis ('ℓ ≈ 5', 'ℓ ≈ 200', 'ℓ ≈ 2000') are hard to read and their mapping to k is not explained in the caption. Add a sentence in the caption or a secondary axis making the k–ℓ correspondence explicit.","section":"Figure 2","authors_acknowledge":null,"authors_acknowledge_p":null},{"comment":"The note that Section 3 was modified after Refs. [33,34] is good practice. Consider moving the substance of the correction into the main text so that readers of the published version see it without consulting the footnote.","section":"Footnote 1","authors_acknowledge":null,"authors_acknowledge_p":null}],"recommendation":"minor_revision","confidential_remarks":"The AIC column error in Table 2 is the kind of slip that should be caught at proof stage, but it is easily fixed and does not affect the primary Δχ² result. The paper is otherwise careful and the correction of the earlier anticorrelation claim is handled well. I would not send this back for a full re-analysis unless the authors cannot show that the fixed-parameter shortcut is harmless; a short decomposition of the χ² improvement by multipole band would be sufficient."},"author_rebuttal":null,"desk_editor":{"model":"selfhosted/deepseek-v4.1-flash","letter":"Worth your time if you care about how r gets extracted from TT, but read Table 2 with a calculator open. The core result holds: a Starobinsky-type step in the inflaton slope, giving ~26% suppression of scalar power below k* ≈ 0.0015/Mpc, fits Planck+WP+BICEP2 better than a constant running by Δχ² ≈ 7.5 for one extra parameter, and the marginalized posteriors in Fig. 4 exclude no-suppression. That is a real, if modest, improvement, and the mechanism is more motivated than the α_s ≈ −0.02 that running requires. The appendix step-function check is the part I trust most: it reproduces the same Δχ² and consistent (k*, β_s) without the ringing, which tells you the gain isn't an artifact of Starobinsky's oscillations. Good, honest robustness work. Section 3 is also clean — the derivation from Eq. 3.3 through 3.4–3.5 showing a primordial ζh correlator drops out of C_ℓ is correct, and footnote 1 openly reverses their own earlier claim after refs. 33–34. That kind of self-correction is rare enough to be worth noting. Now the soft spots. First, Table 2's relative-likelihood column is exp(ΔAIC), not the exp(ΔAIC/2) that Eq. 2.5 defines — ln(2.45×10⁻⁵) = −10.617 exactly. Recomputing from Table 1 gives relative likelihoods of ~0.006 and ~0.062, so suppression beats running ~16:1, not ~385:1. The direction is unchanged; 'strongly favoured' and the abstract's 'considerably better' are oversold by an order of magnitude, and Δχ² = 7.5 is ~2.7σ, not decisive. Second, the MCMC pins every background and nuisance parameter at Planck best-fit values ('for simplicity', Section 2). I think this is probably harmless because the gain sits at ℓ = 20–30 where ω_b, ω_c, θ barely matter, but the paper asserts rather than shows it, and a chi-squared difference this small deserves the full run. Third, r = 0.18 is essentially the BICEP2 input coming back out, not an independent confirmation. None of this is fatal, and the dust caveats are acknowledged. Send it out. It's a competent, clearly written model-comparison paper from a chaotic few weeks, and the referee should ask for the corrected AIC column and a full-parameter run.","headline":"A sensible, well-executed 'what if r really is 0.2' paper whose central comparison survives, but whose AIC table has a factor-of-two slip in the exponent that inflates the headline preference.","tokens_in":15802,"tokens_out":1723,"would_cite":true,"duration_ms":31918,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"pith_extraction":{"msc":[],"pacs":["98.80.Cq","98.70.Vc","98.80.Es"],"model":"selfhosted/deepseek-v4.1-flash","headline":"A sudden 14% change in the inflaton's slope can suppress large-scale scalar power enough to fit a tensor-to-scalar ratio near 0.2 without a large running.","keywords":["cosmic microwave background","inflation","tensor-to-scalar ratio","scalar power spectrum","Starobinsky model","spectral running","BICEP2","Planck satellite"],"falsifier":"A full Planck+WP+BICEP2 Markov chain varying all cosmological and nuisance parameters would test the comparison: if the suppression model's $Delta chi^2$ advantage over constant running falls to the size of its one-parameter penalty, the claim that it fits better does not survive.","tokens_in":15964,"feed_emoji":"📉","tokens_out":8398,"duration_ms":139362,"temperature":0.7,"pith_summary":"The paper asks whether the large tensor-to-scalar ratio reported by BICEP2, $r$ approximately 0.2, can be reconciled with the temperature anisotropies without invoking a running of the scalar spectral index much larger than slow roll predicts. Its proposal is that scalar power is suppressed at large scales by a sudden change in the slope of the inflaton potential, so the temperature spectrum loses power where tensors contribute most while the B-mode signal remains. Using a Starobinsky step model, the authors find a best fit with a 14% change in slope, about 26% suppression below $k_*$ approximately $0.0015$ per Mpc, and $r=0.18$; the effective $chi^2$ improves by about 7.5 over constant running at the cost of one parameter, and the AIC also prefers the suppression model. If this holds, a high-scale inflationary signal in B modes need not be accompanied by an exotic running. The paper also checks whether a scalar-tensor anticorrelation could do the same job and concludes that it cannot, because it changes off-diagonal correlations but not the temperature power spectrum.","feed_headline":"A kink in the inflaton's slope can rescue a large r","feed_subtitle":"A 26 percent drop in large-scale scalar power fits the temperature data better than running.","key_machinery":"The load-bearing mechanism is the Starobinsky piecewise-linear potential, $V(varphi)=V_0+C_+(varphi-varphi_*)$ for $varphi$ greater than or equal to $varphi_*$ and $V(varphi)=V_0+C_-(varphi-varphi_*)$ for $varphi$ less than or equal to $varphi_*$ with $C_- < C_+$. Its scalar power spectrum has a large-scale plateau suppressed by $(1- Delta)^2$, a small ringing from the sharp transition, and the usual tilt at small scales; the paper fits the full template rather than a bare step. The fit uses modified CAMB and CosmoMC, adding $k_*$ and $Delta$ to a reduced parameter set. For the anticorrelation section, the machinery is a scale-invariant $langle zeta h_{ij} rangle$ correlator projected onto transverse-traceless modes; the induced $C_{ell m ell+2m}$ and $C_{ell m ell m}$ terms average to zero in $C_ell$.","core_discovery":"The central claim is that a step-like suppression of the primordial scalar power at large scales can fit the Planck temperature data, WMAP9 polarization, and BICEP2 better than a constant running spectral index while keeping $r$ approximately 0.2. In the Starobinsky model the inflaton rolls from a steeper to a shallower slope; modes that leave the horizon before the step have their power reduced by $(1- Delta)^2$, while modes after it are unchanged. With $tau$, $n_s$, $ln(10^{10}A)$, $r$, $k_*$, and $Delta$ free and the other parameters fixed, the best fit is $k_* = 0.0015$ per Mpc, $Delta = 0.14$ (about 26% suppression), and $r=0.18$, with effective $chi^2 = 9838.87$ versus $9846.43$ for constant running, a $Delta chi^2$ approximately 7.5 for one extra parameter. The paper also shows that a primordial scalar-tensor anticorrelation produces off-diagonal $C_{ell m ell' m'}$ but leaves $C_ell$ unchanged, so it cannot lower the TT spectrum used to bound $r$; the authors note that this conclusion was revised from an earlier version after [33,34].","pith_inferences":["Editorial extension: a full-parameter likelihood run is the obvious next check; if the $Delta chi^2$ advantage shrinks when Planck's background and nuisance parameters vary, the suppression case would rest on the AIC penalty rather than on the fit itself.","The step's ringing, though minor here, leaves a calculable oscillatory imprint in the TT spectrum around the transition; a dedicated search at low $ell$ could distinguish a sharp Starobinsky step from a smooth sound-speed suppression.","If dust cleaning lowers BICEP2's $r$ below about 0.1, the original tension disappears, but the same large-scale suppression could still be tested as an independent explanation of the low-$ell$ TT deficit.","The off-diagonal correlations induced by a scalar-tensor correlator are a possible target for statistical-anisotropy searches, even though they cannot lower the temperature power spectrum."],"forward_implications":["A confirmed $r$ approximately 0.2 would not force a running $alpha_s$ approximately $-0.02$; a step in the potential can absorb the tension while the inflaton stays in slow roll before and after the transition.","The preferred suppression is concrete: about 26% less scalar power below $k_*$ approximately $1.5$ times $10^{-3}$ per Mpc, corresponding to the largest observable scales and the low-$ell$ temperature points.","Because the tensor power is continuous across the step, the model keeps a B-mode signal at $r$ approximately 0.18 while lowering only the large-scale scalar TT contribution.","A scalar-tensor anticorrelation is ruled out as a route to lowering the TT bound on $r$, since it changes off-diagonal correlations but not $C_ell$.","A ringing-free step-function parametrization gives consistent $k_*$ and $beta_s$, so the fit does not depend on the oscillatory part of the Starobinsky template."],"supporting_citations":[{"why":"Supplies the BICEP2 B-mode detection and the $r$ approximately 0.2 value the paper tries to reconcile.","marker":"[1]"},{"why":"Supplies the Planck temperature bound $r<0.11$ and the running comparison the paper improves on.","marker":"[3]"},{"why":"Supplies the piecewise-linear potential and the scalar power template with a step and ringing.","marker":"[28]"},{"why":"Supplies the explicit form of the Starobinsky power spectrum used in the fit.","marker":"[40]"},{"why":"Shows that scalar-tensor correlation leaves the temperature power spectrum unchanged, grounding the negative result.","marker":"[33]"},{"why":"Independently reaches the same conclusion about the anticorrelation's no-effect on $C_ell$.","marker":"[34]"},{"why":"Supplies CAMB, modified here to compute scalar plus tensor TT and BB spectra.","marker":"[41]"},{"why":"Supplies CosmoMC, modified here to sample the extra parameters $k_*$ and $Delta$.","marker":"[42]"},{"why":"Supplies the Planck base-$r$ best-fit values at which the other cosmological and nuisance parameters are held fixed.","marker":"[43]"}],"fun_headline_variants":["A 14 percent inflaton speed change fits large r better than running","Step-like scalar suppression beats constant running at r = 0.18","Anticorrelation leaves TT unchanged and cannot ease large-r bounds","Revised null result: anticorrelation cannot lower the TT spectrum","One extra parameter improves chi-squared by 7.5 over running"],"cache_read_input_tokens":0,"weakest_assumption_plain":"The reconciliation rests on the BICEP2 B-mode being primordial and on a model comparison in which every cosmological and nuisance parameter except $tau$, $n_s$, $ln(10^{10}A)$, $r$, $k_*$, and $Delta$ is held fixed at the Planck base-$r$ best fit.","fun_headline_variants_meta":{"raw":{"variants":["A 14 percent inflaton speed change fits large r better than running","Step-like scalar suppression beats constant running at r = 0.18","Anticorrelation leaves TT unchanged and cannot ease large-r bounds","Revised null result: anticorrelation cannot lower the TT spectrum","One extra parameter improves chi-squared by 7.5 over running"]},"model":"selfhosted/deepseek-v4.1-flash","effort":"low","attempts":1,"cost_usd":0.0,"raw_usage":{"total_tokens":2235,"prompt_tokens":1105,"completion_tokens":1130,"prompt_tokens_details":null,"completion_tokens_details":{"reasoning_tokens":1053}},"tokens_in":1105,"tokens_out":1130,"duration_ms":20993,"image_pages":[],"temperature":1.0,"reasoning_budget":50,"reasoning_tokens":1053,"transport_provider":"selfhosted","cache_read_input_tokens":0,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-11-10T08:12:14.842286+00:00","model_set":{"reader":"selfhosted/deepseek-v4.1-flash"},"falsifier":"A full Planck+WP+BICEP2 Markov chain varying all cosmological and nuisance parameters would test the comparison: if the suppression model's $Delta chi^2$ advantage over constant running falls to the size of its one-parameter penalty, the claim that it fits better does not survive.","supporting_citations":[{"cited_title":"BICEP2 I: Detection Of B-mode Polarization at Degree Angular Scales","cited_arxiv_id":"1403.3985","evidence_quote":"Supplies the BICEP2 B-mode detection and the $r$ approximately 0.2 value the paper tries to reconcile."},{"cited_title":"Planck 2013 results. XXII. Constraints on inflation","cited_arxiv_id":"1303.5082","evidence_quote":"Supplies the Planck temperature bound $r<0.11$ and the running comparison the paper improves on."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the piecewise-linear potential and the scalar power template with a step and ringing."},{"cited_title":"The scalar bi-spectrum in the Starobinsky model: The equilateral case","cited_arxiv_id":"1109.5838","evidence_quote":"Supplies the explicit form of the Starobinsky power spectrum used in the fit."},{"cited_title":"Scalar-tensor correlations and large-scale power suppression","cited_arxiv_id":"1404.4866","evidence_quote":"Shows that scalar-tensor correlation leaves the temperature power spectrum unchanged, grounding the negative result."},{"cited_title":"Scalar-Tensor Anti-Cross-Correlation does not Help to Reconcile the Tension between BICEP2 and Planck","cited_arxiv_id":"1404.5112","evidence_quote":"Independently reaches the same conclusion about the anticorrelation's no-effect on $C_ell$."},{"cited_title":"Efficient Computation of CMB anisotropies in closed FRW models","cited_arxiv_id":"astro-ph/9911177","evidence_quote":"Supplies CAMB, modified here to compute scalar plus tensor TT and BB spectra."},{"cited_title":"Cosmological parameters from CMB and other data: a Monte-Carlo approach","cited_arxiv_id":"astro-ph/0205436","evidence_quote":"Supplies CosmoMC, modified here to sample the extra parameters $k_*$ and $Delta$."},{"cited_title":"Planck 2013 results. XVI. Cosmological parameters","cited_arxiv_id":"1303.5076","evidence_quote":"Supplies the Planck base-$r$ best-fit values at which the other cosmological and nuisance parameters are held fixed."}],"review_version":1}