{"id":"79e29fe0-9436-4a36-ad3b-fb211313d9c2","arxiv_id":"1908.10637","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":1,"one_line_summary":"The low large-scale CMB variance, concentrated away from the Galactic plane, remains anomalous at about 2.8 to 3 sigma even after random rotations of the sky are taken into account.","lead":"This paper tests whether the CMB's surprisingly low large-scale temperature fluctuations are linked to the orientation of our Galaxy's plane, by rotating the Planck sky map and its simulations. The anomaly persists after accounting for all possible Galactic orientations, at roughly 3 sigma, which strengthens the case that the effect is real rather than a geometric accident.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Rotation test is sound, but the headline 2.8–3.1 sigma values are quoted at Ext30, the most extreme of five masks, without correcting for the look-elsewhere over mask size and estimator choice; the p-values should be re-evaluated as global statistics.","rationale":"The reader's conditional verdict is appropriate, but the most load-bearing concern is not the one named as the weakest assumption. The null-model issue is real but conservative: Planck 2018 best-fit parameters are estimated from the same data, so the simulated low-l variance is pulled toward the observed low value, making the quoted p-values smaller rather than larger; selecting a high-l-only fit would likely leave the anomaly unchanged or strengthen it. The ensemble-1 analysis further addresses the 'given low variance' version of the null. The rotation look-elsewhere itself is handled correctly: the data's LTP_c is the fraction of its own rotations with lower variance, and the comparison to simulations is a calibration of the uniform distribution. The genuine unaddressed multiplicity is the choice of Ext30 as the headline mask after observing a monotonic trend across five masks, plus the choice of two estimators. A global statistic over masks and estimators is the direct check. The paper's qualitative conclusion likely survives, but the quoted 2.8–3.1 sigma should be presented as conditional on mask selection, not as a global significance. Therefore the reader's CONDITIONAL verdict stands unchanged, with the mask-selection caveat made explicit.","tokens_in":17116,"tokens_out":13645,"duration_ms":156328,"concrete_test":"Compute, for each of the 10^3 ensemble-0 simulations, the five per-mask LTP_i values and the four per-mask r_i values using the same 1000 rotations per map. Define global statistics G_LTP = min over masks of LTP_i and G_r = max over masks of r_i (or min over masks of UTP). Evaluate G_LTP and G_r for the Commander and SMICA maps and compare them with the simulation distributions of G_LTP and G_r. If the global tail probabilities stay below 1–2%, the Ext30 significance is robust to mask selection; if they rise above 5%, the headline 2.8–3.1 sigma is largely a product of selecting the most extreme mask.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The rotation-based look-elsewhere test is well designed: LTP_c for the data is the fraction of 1000 rotations with variance below the true orientation, so it is already a rotation look-elsewhere p-value; the comparison of LTP_c to the distribution of LTP_i is a calibration and gives the same 0.5–0.7% level. The load-bearing problem is that the headline 2.8–3.1 sigma figures are quoted at Ext30, the most extreme of the five masks, after the monotonic trend has been seen. The paper never applies a look-elsewhere correction over the five masks or over the LTP/r estimators. Under the null, the minimum LTP across five masks is stochastically smaller than the LTP at any fixed mask, so the global tail probability for the data will be larger than the reported 0.5% (LTP) and 0.2% (r). A simple Bonferroni bound gives about 2.5% and 1% respectively, and the true global p after accounting for correlations among masks and estimators could be a few percent. This does not refute the qualitative claim of stability against rotations, but it means the quantitative sigma values are post-selection. The reader's null-model concern is second-order: the Planck best-fit is pulled toward the data's low variance, so the reported p-values are if anything conservative, and the ensemble-1 analysis already conditions on the observed low variance.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper investigates whether the well-known lack-of-power anomaly at large angular scales in the CMB is specifically associated with high Galactic latitude, and whether this association could be explained by the particular orientation of the Galactic plane. Using Planck 2018 Commander and SMICA temperature maps, five Galactic masks, a quadratic maximum-likelihood power-spectrum estimator, and 10^5 Lambda-CDM simulations, the authors first reproduce the known result that the variance lower-tail probability decreases as the Galactic mask grows, reaching about 0.3-0.5% for the most extended Ext30 mask. They then apply random rotations to the maps and define two estimators: the LTP estimator, which measures how often a random rotation yields a lower variance than the unrotated map, and the r estimator, which measures the variance drop relative to the largest drop across rotations. For the Ext30 mask they report a 0.5% LTP value (about 2.8 sigma) for Commander and a 0.2% upper-tail probability (about 3.1 sigma) for the r estimator. The analysis is repeated with a low-variance constrained ensemble (ensemble 1) and gives similar results, and robustness checks with different thresholds and with Planck 2015 data are provided in the appendices.","tokens_in":17323,"tokens_out":6172,"duration_ms":63118,"significance":"If the result holds, it is a useful and interesting contribution to the CMB-anomaly literature: it shows that the high-latitude lack of power is not simply a consequence of the Galactic plane orientation being special, and it provides a concrete rotation-based look-elsewhere procedure. The Monte Carlo design is careful and extensive: 10^5 Lambda-CDM realisations, QML C_ell estimation, two component-separation maps, five masks, a low-variance constrained ensemble, a validation of the rotation generator in Appendix A, threshold-dependence checks in Appendix C, and a comparison with the Planck 2015 release in Appendix B. The paper is honest about the exploratory nature of several choices. The main weakness is that the headline significances are quoted for the most extreme of five masks and for two estimators without a global correction for the mask and estimator scan, so the quantitative claim as stated is not yet fully supported.","major_comments":[{"comment":"The headline significances (2.8 sigma from the LTP estimator and 3.1 sigma from the r estimator) are quoted at the Ext30 mask, which is the most extreme of the five masks in Table 1 and is the mask where the monotonic trend in Figure 4 is largest. No correction is applied for scanning over the five masks or over the two estimators. Under the null, the minimum LTP across masks is stochastically smaller than the LTP at any fixed mask, so the reported 0.5% and 0.2% values are post-selection. A simple Bonferroni correction over five masks gives at least roughly 2.5% (LTP) and 1% (r), and the true global p-value after accounting for correlations among masks and estimators could be a few percent. This is load-bearing because the abstract and conclusions explicitly quote the 2.8-3.1 sigma levels as the evidence that the anomaly survives the look-elsewhere effect. Please provide a global statistic, for example the distribution over simulations of the minimum LTP (or maximum r) across masks and estimators, and report the fraction of simulations that exceed the observed value, or explicitly re-label the quoted numbers as mask-dependent exploratory findings.","section":"Section 3.1.1 and Section 3.1.2, Tables 2 and 3"},{"comment":"The similarity of the ensemble 1 results is presented as supporting the main conclusion, but ensemble 1 was selected after the fact to reproduce the observed low variance. This is a conditional test, not an independent confirmation of the anomaly. Its valid and useful role is to show that a low total variance in the standard mask is not by itself sufficient to generate the high-latitude trend; however, the text should state this caveat explicitly rather than implying that the ensemble 1 analysis independently validates the 2.8-3.1 sigma result. The distinction matters because the ensemble 1 p-values inherit the same mask-selection issue identified above and are also conditioned on the data variance.","section":"Section 4 and Tables 5-6"}],"minor_comments":[{"comment":"The word 'bizzarre' should be 'bizarre', and in Section 5 'simular' should be 'similar'.","section":"Abstract and Section 1"},{"comment":"The text states that 'only 5 maps out of 10^5 have a LTP... smaller than the Planck Commander data', but Section 3.1.1 and Table 2 report 5 out of 10^3 (0.5%). This is a numerical inconsistency that should be corrected.","section":"Section 5, final paragraph"},{"comment":"The caption of Figure 10 appears to swap 'Left panel' and 'Right panel' relative to the text: the panel showing the LTP versus sky fraction is on the left, while the panels with l_min = 3 and l_min = 4 are on the right. Please fix the caption.","section":"Figure 10 caption and Section 4"},{"comment":"The denominator of the r-estimator is described as the maximum over rotations, but the unrotated case is included only in a footnote. The main text should explicitly state that the denominator includes the unrotated orientation, since the sign and the normalization of r depend on this convention.","section":"Equation (3.2)"},{"comment":"The null hypothesis is the Planck 2018 best-fit Lambda-CDM model, which was derived from the same data being tested. The paper should state explicitly that all p-values are conditional on this model and briefly comment on the expected direction of any bias; a one-sentence caveat in Section 2.2 or Section 5 would suffice.","section":"Section 2.2"},{"comment":"The left panel uses the notation V_c^{(rot)} < V_c, which is not defined in the table caption; please define the notation there.","section":"Table 2"}],"recommendation":"major_revision","confidential_remarks":"The paper is within scope for JCAP and the rotation-based approach is a genuine methodological step forward. The key issue for the editor is the post-selection nature of the Ext30 significance: the authors should be asked to provide multi-mask and multi-estimator global p-values, or to tone down the quantitative claims accordingly. The self-citation pattern is noticeable but not excessive for this specialized subfield."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Short version: the paper's real addition is the rotation-based analysis—the LTP and r estimators—and the claim that the high-latitude low-variance anomaly is not an artifact of the particular orientation of the Galactic plane. That claim is credible. The rotation look-elsewhere is properly handled: LTP_c for the data is the fraction of 1000 rotations with variance below the true orientation, and comparing that LTP_c to the distribution of LTP_i is a valid calibration. The r-estimator, which normalizes the variance drop by the maximum drop across rotations, is a sensible way to test the trend.\n\nWhat the paper does well is solid Monte Carlo machinery: 10^5 Lambda-CDM sims, QML spectra, five masks, two maps (Commander and SMICA), plus robustness checks with a low-variance constrained ensemble, threshold changes, and a 2015 data check. The finding that the trend persists when conditioning on the observed low total variance (ensemble 1) is a good control—it shows the high-latitude behavior is not just a consequence of global low power.\n\nSoft spots, in proportion. The headline significances are quoted at Ext30, the most extended mask, after the monotonic trend has been seen. No look-elsewhere correction is applied over the five masks or over the LTP/r estimators. A Bonferroni bound over five masks would push the LTP p-value to a few percent and the r p-value to about one percent; the true global p after accounting for mask and estimator correlations could be a few percent. That does not refute the qualitative conclusion—the anomaly is still more likely than not to be real—but it means '2.8–3.1 sigma' is post-selection and should not be read as a global significance. The authors discuss the a posteriori orientation but not the mask selection. Minor: the null is the Planck 2018 best-fit derived from the same data, so p-values are conditional on that model. As the stress-test notes, this is if anything conservative, because the best-fit is pulled toward the data's low variance.\n\nBottom line: this is a careful, honest paper that adds a genuinely new test. The qualitative claim survives scrutiny; the quantitative significance should be reworded or corrected. It deserves peer review. The main required revision is an explicit global p-value across masks/estimators, or a toned-down headline. I'd bring it to reading group, though it's not a blockbuster.\n\nRecommendation: send to referees.","headline":"A careful MC study of the CMB low-variance anomaly that adds a rotation look-elsewhere test; the central qualitative claim holds, but the headline 2.8–3.1 sigma are quoted at the most extreme mask without a global mask/estimator correction.","tokens_in":17957,"tokens_out":1906,"would_cite":true,"duration_ms":20133,"reading_group":"yes","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 claims that the CMB's high-latitude low-variance (lack-of-power) anomaly survives random rotations of the Galactic orientation, at about 2.8 to 3.1 sigma, so it is not a mask-selection fluke.","keywords":["cosmic microwave background","CMB anomalies","lack of power","low variance","Galactic mask","look-elsewhere effect","angular power spectrum","Planck 2018"],"falsifier":"Recompute the lower-tail probability and r-estimator using a full-sky CMB map from an independent experiment, or a map in which the Galactic-plane gap is filled by a validated inpainting scheme rather than masked; if the high-latitude variance is no longer in the lowest 0.5 percent of rotated Lambda CDM realisations, the claimed orientation-stable anomaly is a mask artifact.","tokens_in":16830,"feed_emoji":"🌌","tokens_out":6315,"duration_ms":63760,"temperature":0.7,"pith_summary":"This paper asks whether the CMB's 'lack of power' anomaly—the observation that large-scale temperature fluctuations are weaker than the standard cosmological model predicts—is an artifact of choosing the orientation of our Galaxy as a reference. It reproduces the known trend that the anomaly becomes stronger as more of the Galactic plane is masked out, reaching about 3 sigma in the most aggressive high-latitude mask. The new step is to rotate data and simulated maps randomly, which mimics a look-elsewhere effect across all possible orientations of the Galaxy. The claim is that the high-latitude low variance remains anomalous after this resampling: the lower-tail probability is 0.5 percent for one estimator and 0.2 percent for the trend estimator, corresponding to roughly 2.8 and 3.1 sigma. If correct, the orientation of our Galaxy is not what is producing the anomaly, and explanations must account for why the deficit is strongest away from the Galactic plane.","feed_headline":"2.8-sigma CMB power deficit survives sky rotations","feed_subtitle":"Rotating the Galaxy plane does not explain the lack of power at high latitude; the anomaly is intrinsic to the sky.","key_machinery":"The central object is the map variance $V = \\sum_{\\ell=2}^{29} \\frac{2\\ell+1}{4\\pi} C_\\ell$, estimated with the Quadratic Maximum Likelihood angular power-spectrum estimator. The new machinery is rotation-resampling: each map is rotated many times via Wigner rotation matrices applied to spherical-harmonic coefficients, and the distribution of $V$ across rotations defines a lower-tail probability for each map. The $r$-estimator $r=(V_{\\rm std}-V_{\\rm mask})/\\max_j(V_{\\rm std}^{(j)}-V_{\\rm mask}^{(j)})$ compares the observed variance drop between the standard and extended masks with the largest drop among rotations, separating the anomaly's dependence on the mask's sky fraction from its dependence on orientation.","core_discovery":"We show that the lack of power at high Galactic latitude is substantially stable against the look-elsewhere effect induced by random rotations of the Galactic orientation. The variance of the temperature map, computed from the angular power spectrum up to l=29, decreases as more sky around the Galactic plane is masked; random rotations of the same maps show that the observed decrease is near the extreme of what rotations can produce, with r around 0.88-0.90 and an upper-tail probability of 0.2 percent. The lower-tail probability estimator gives 0.5 percent for Commander and 1.3 percent for SMICA in the most aggressive mask, meaning only 5 (or 13) of 1000 Lambda CDM realisations have a more anomalous orientation. Repeating the analysis with simulations constrained to have the observed low variance gives essentially the same result, so the anomaly is not merely a consequence of the map having low power to begin with.","pith_inferences":["A natural testable extension is to apply the same rotation-resampling to other CMB anomalies, such as hemispherical asymmetry or parity asymmetry, to see whether their significance is also orientation-stable.","If the anomaly is genuinely orientation-independent, then masks centred on randomly chosen sky directions should not reproduce the same high-latitude deficit unless they exclude the actual Galactic plane; this could be checked directly.","The result sharpens the challenge for models that try to explain the low-l power with a localized physical feature: such a feature would have to avoid the Galactic plane rather than merely suppress power globally.","Combining rotation-resampling with a validated full-sky inpainting method could test whether the mask itself, through mode coupling, contributes to the apparent anomaly."],"forward_implications":["The high-latitude low-variance anomaly is not a selection effect of Galactic orientation; any explanation must produce a deficit that is intrinsically stronger away from the Galactic plane.","The decreasing variance trend with mask size is itself anomalous against rotations, with the observed maps falling in the upper 0.2 percent of the r-estimator distribution, so the anisotropy of the anomaly is part of the signal.","The anomaly persists when simulations are constrained to have the same low total variance as the data, ruling out the interpretation that a generally low-variance sky naturally produces this latitude trend.","The effect is dominated by the quadrupole and octupole, and at low latitudes by odd multipoles, pointing to the lowest multipoles as the locus of the signal."],"supporting_citations":[{"why":"Established the mask dependence of the low-variance anomaly that this paper extends.","marker":"[15]"},{"why":"Showed the anomaly is dominated by odd multipoles, a result this paper reproduces and builds on.","marker":"[16]"},{"why":"Previous Planck 2015 detection of low variance at 2-3 sigma, the anomaly under study.","marker":"[5]"},{"why":"Planck 2018 results used as the fiducial model and data context.","marker":"[6]"},{"why":"Supplies the standard mask whose extensions define the analysed masks.","marker":"[20]"},{"why":"Provides the Commander and SMICA 2018 temperature maps used as data.","marker":"[21]"},{"why":"Provides the pixelisation and harmonic rotation machinery used for random rotations.","marker":"[23]"},{"why":"Supplies the Quadratic Maximum Likelihood angular power spectrum estimator used to build variances.","marker":"[24]"}],"fun_headline_variants":["Galactic tilt can't explain CMB power deficit","CMB large-scale power loss survives sky rotations","CMB anomaly not a look-elsewhere artifact, rotations fail","High-latitude CMB variance deficit is intrinsic to the sky"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The whole significance calculation assumes that the standard cosmological model fitted to the Planck data is the right description of the true large-scale sky; if that fit already absorbed part of the missing power, the quoted probabilities would shift.","fun_headline_variants_meta":{"raw":{"variants":["Galactic tilt can't explain CMB power deficit","CMB large-scale power loss survives sky rotations","CMB anomaly not a look-elsewhere artifact, rotations fail","High-latitude CMB variance deficit is intrinsic to the sky"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000228,"raw_usage":{"total_tokens":1538,"prompt_tokens":1070,"completion_tokens":468,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":686,"completion_tokens_details":{"reasoning_tokens":401}},"tokens_in":686,"tokens_out":468,"duration_ms":5395,"temperature":1.0,"reasoning_tokens":401,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T10:38:12.730275+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Recompute the lower-tail probability and r-estimator using a full-sky CMB map from an independent experiment, or a map in which the Galactic-plane gap is filled by a validated inpainting scheme rather than masked; if the high-latitude variance is no longer in the lowest 0.5 percent of rotated Lambda CDM realisations, the claimed orientation-stable anomaly is a mask artifact.","supporting_citations":[{"cited_title":"The Evens and Odds of CMB Anomalies","cited_arxiv_id":"1712.03288","evidence_quote":"Showed the anomaly is dominated by odd multipoles, a result this paper reproduces and builds on."}],"review_version":1}