{"id":"8e77278b-341b-4c5e-9b27-69be384400e3","arxiv_id":"2506.22795","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":4.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":4,"one_line_summary":"Using Planck PR4, nine low multipoles show anomalous anisotropy with a cumulative probability of about 0.3 percent, while quadrupole-octopole alignment is only marginal and no collective alignment is found.","lead":"This paper tests whether the largest-scale patterns in the cosmic microwave background violate the rule that the sky looks the same in every direction, using the final Planck data release. It finds a persistent hint at about the 3-sigma level that several low multipoles are intrinsically anisotropic, while other alignments are weaker.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"0.28% cumulative significance depends on l=6, whose p-value straddles the 0.05 threshold across the two simulation ensembles.","rationale":"The paper's central quantitative result is the cumulative probability Pcuml ≈ 0.0028 for observing 9 of 60 multipoles with Power-entropy p≤0.05. This requires two things: the per-mode p-values from simulations must be reliable, and the count of modes below threshold must be stable. The l=6 entry in Table 1 violates the second condition: its p-value is 0.03 using the 100 Planck-like simulations but 0.0534 using the 5000 ideal realizations. Since the ideal realizations are not described as being masked and inpainted, part of the discrepancy may stem from different sky treatments rather than statistical scatter. Excluding l=6 changes the headline from ~3σ to ~2.6σ and Pcuml from 0.0028 to ~0.009. The reader's weakest assumption about inpainting bias is correct in spirit, but the sharper issue is that the 0.05 threshold is crossed by a single mode whose inclusion is not robust to the simulation pipeline. The paper's supporting checks—agreement of the two 95% CL curves and the subset relation between planar and axial anomalies—do not address this threshold sensitivity. A targeted recomputation after processing the ideal realizations through the same inpainting pipeline, and repeating the data analysis with alternative masks, would settle whether the 0.3% claim holds. If l=6 moves across the threshold, the conclusion should be downgraded to a ~2.5σ hint; if it stays, the inpainting-bias concern is alleviated. Neither outcome changes the overall verdict from CONDITIONAL, so the reader's recommendation stands.","tokens_in":13590,"tokens_out":6984,"duration_ms":71718,"concrete_test":"Take the 5000 ideal CMB realizations, apply the same extended WMAP kp8 mask (fsky≈0.92) and iSAP sparse inpainting procedure used for the data and the 100 FFP simulations (Section 3, Appendix A), and recompute the Power-entropy p-values and k* for Eq. (7). Also repeat the data analysis with the more conservative Planck common inpainting mask (fsky≈0.979) and the aggressive common mask (fsky≈0.78). If l=6 is excluded or included depending on mask/inpainting choice, then the 0.3% cumulative significance is a pipeline artifact rather than a robust property of the data.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Eq. (7) yields Pcuml ≈ 0.0028 only because k*=9 modes fall below p≤0.05 (Table 1), and that count is not robust. The l=6 mode has p=0.03 from the 100 Planck PR4 simulations but p=0.0534 from the 5000 ideal CMB realizations—just past the threshold. Removing it gives k*=8 and Pcuml ≈ 0.009, closer to 2.6σ than 3σ. The two ensembles are not directly comparable: Appendix A states that the 100 Planck simulations are downgraded and inpainted exactly as the data, but the 5000 ideal realizations are described only as generated at Nside=256 with 40′ beam; no masking/inpainting step is mentioned for them. The paper's claim that the ensembles 'compare fairly well' hides this threshold-level discrepancy. Moreover, with only 100 simulations a p-value of 0.03 means 3 of 100 realizations fall below the observed S(l); the sampling uncertainty (≈±1.7 events) alone can move l=6 across the 0.05 line. Thus the central 0.3% rejection of statistical isotropy rests on a single borderline multipole whose status depends on an unvalidated simulation treatment.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper applies the Power tensor and Alignment tensor statistics to the Planck PR4 NPIPE Commander temperature map, restricted to multipoles l=2 to 61, and compares the observed statistics with 100 Planck FFP simulations and 5000 ideal CMB realizations generated from a best-fit theoretical power spectrum. The central result is that 9 of the 60 multipoles have Power entropy p-values at or below 0.05, giving a binomial cumulative probability Pcuml ≈ 0.0028 (claimed as close to 3-sigma significance), which the authors interpret as evidence for a significant number of intrinsically anisotropic low multipoles. Additional results concern planarity (no independently planar modes beyond those already anisotropic), alignment of l=3,44,59,61 with the quadrupole, and the absence of collective alignment in the range l=2-61 according to the Alignment tensor.","tokens_in":13834,"tokens_out":6812,"duration_ms":61814,"significance":"The Power tensor is an established public statistic, and the paper provides a useful update using the latest Planck PR4 map. The analysis pipeline is described in detail, and the use of two simulation ensembles gives a degree of cross-checking. The headline result is falsifiable and the binomial global test is a reasonable way to assess the multiplicity of low-l anomalies. However, the significance is borderline and is sensitive to the choice of reference simulations and to the inpainting/mask treatment, so the robustness of the central claim needs to be demonstrated.","major_comments":[{"comment":"The cumulative probability Pcuml ≈ 0.0028 depends on counting k* = 9 modes with p ≤ 0.05; this count is not robust because the l=6 mode has p = 0.03 from the 100 PR4 simulations but p = 0.0534 from the 5000 ideal realizations, straddling the threshold. If the latter ensemble is used as the reference, k* becomes 8 and Pcuml increases to roughly 0.01, reducing the claimed significance from ~3σ to ~2.5σ. The paper must either justify why the 100 PR4 simulations are the primary reference or show that the discrepancy for l=6 is within sampling uncertainty.","section":"§5, Eq. (7), Table 1"},{"comment":"The two simulation ensembles are not processed identically: the 100 PR4 simulations are downgraded and inpainted exactly as the data, while the 5000 ideal CMB realizations are generated at Nside=256 with a 40′ beam and no masking or inpainting step is described. The p-values from the ideal set are therefore not a clean cross-check for a masked and inpainted data map. The authors should process the ideal realizations through the same mask/inpainting pipeline, or explicitly restrict the cross-check to statistics that are insensitive to inpainting, and explain why the l=6 p-value differs across ensembles.","section":"§3, Appendix A"},{"comment":"The assertion that iSAP sparse inpainting 'preserves the statistical properties with the rest of the unmasked sky' is not validated for the chosen mask (fsky ≈ 0.92) and multipole range. The mask is selected after testing several options and rejecting those that produce visible discontinuities, which is a data-dependent choice that is not accounted for in the p-values. The authors should validate the inpainting assumption, for example by comparing Power tensor distributions from inpainted ideal simulations with those from full-sky ideal simulations, and should test the sensitivity of the anomalous-mode list to the mask choice.","section":"§3, Appendix A"}],"minor_comments":[{"comment":"The 'cumulative probability of 0.3%' is a rounded version of 0.0028 ≈ 0.28%; please report the exact value (e.g., 0.0028) alongside the rounded figure.","section":"Abstract, §5"},{"comment":"For l=30, the p-value from the 100 PR4 simulations is reported as '<0.01,' which is a coarse upper limit; consider reporting the exact binomial confidence interval or the fraction of simulations below the observed value.","section":"§4.1, Table 1"},{"comment":"The quadrupole-octopole alignment p-values (0.08 and 0.035) are described as 'marginally anomalous at about 2σ level,' but p=0.08 corresponds to roughly 1.4σ for a one-sided test; the language should be consistent with the quoted p-values.","section":"§4.2"},{"comment":"The sky fraction of the mask is stated as both fsky ≈ 0.92 and fsky ≈ 0.9; please make the values consistent.","section":"Appendix A"},{"comment":"The histograms for the Alignment entropy and the smallest eigenvalue use similar colors; add distinguishing line styles or labels for clarity.","section":"Fig. 5"},{"comment":"The term 'intrinsically anisotropic' is used for multipoles with small Power entropy p-values; this conflates a statistical statement about the data with a physical property. Suggest using 'significantly anisotropic' or 'anomalously anisotropic' instead.","section":"Throughout"}],"recommendation":"major_revision","confidential_remarks":"The paper addresses a well-known topic with a standard statistic. The main risk is the borderline significance of the headline result, which depends on a single multipole and on the treatment of the simulation ensembles. The authors should be asked to process the ideal simulations through the same mask/inpainting pipeline and to validate the inpainting assumption; the mask-selection procedure should also be justified as not introducing a posteriori bias. If these issues are resolved, the paper could be acceptable."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Aluri and Patel apply the established Power tensor and Alignment tensor statistics to the Planck PR4 Commander temperature map for l=2–61. That is new as an application, not as a method. The genuinely new numbers are the per-multipole p-values, the list of nine anomalous multipoles, the cumulative binomial probability of 0.3%, the quadrupole–octopole alignment angle, and the claim that the collective Alignment-tensor axis sits near several low-l axes. The paper does well on transparency: dipole removal, downgrading, mask construction, and inpainting are described in enough detail to reproduce, and they run two simulation sets. That is real care.\n\nThe soft spots are concentrated in the central claim. The cumulative 0.3% rejection of statistical isotropy comes from counting k*=9 multipoles with p≤0.05. That count is not robust. l=6 has p=0.03 from the 100 PR4 simulations but p=0.0534 from their 5000 ideal realizations. Removing it drops k* to 8 and P_cuml to about 0.009 (2.6σ). The two ensembles are not directly comparable: the FFP simulations are masked and inpainted exactly like the data, while the ideal realizations are just generated at Nside=256 with a 40′ beam, with no masking or inpainting step described. The paper says the ensembles compare fairly well, but that hides threshold-level discrepancies. With only 100 FFP simulations, p=0.03 means three events; sampling uncertainty alone can move l=6 across the line. The same ensemble disagreement shows up in the planarity analysis: the smallest-eigenvalue p-values are ≤0.05 for six multipoles in the FFP set, but all are >0.1 in the 5000 ideal set. The paper concludes there are no truly planar modes, which may be right, but the evidence is contradictory rather than convergent.\n\nThe quadrupole–octopole alignment is also marginal: p=0.08 versus 0.035, and the paper acknowledges this. The AT-PEV proximity to individual axes in Table 4 is interesting but has no uncertainties attached and is not supported by the AT significance tests, which come out unremarkable (p≈0.2). Treat that as a curiosity, not a result.\n\nThe work is worth taking seriously. The method and data are public, and the paper is honest about most of its choices. But the headline significance is fragile in a specific, identifiable way, and the paper would benefit from a robustness section that varies the p-value threshold, processes the 5000 ideal maps through the same masking and inpainting, and reports the l=6 sensitivity explicitly. I recommend peer review with expectations of revision, not rejection. The core observation that a handful of low multipoles look anisotropic is not new, but the PR4 numbers are a useful update, and the 0.3% claim needs to be either shored up or qualified before it becomes a firm statement in the literature.","headline":"A careful PR4 update of the Power-tensor anomalies whose headline 0.3% significance is one borderline multipole away from being a 0.9% result.","tokens_in":14408,"tokens_out":3593,"would_cite":true,"duration_ms":36567,"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":"Using the latest Planck PR4 temperature map, the paper finds that nine of the first sixty CMB multipoles are intrinsically anisotropic, a deviation from statistical isotropy with a cumulative probability of about 0.3% — close to 3-sigma…","keywords":["cosmic microwave background","statistical isotropy","CMB anomalies","low multipoles","Power tensor","Planck PR4","multipole alignment","quadrupole-octopole alignment"],"falsifier":"Run many statistically isotropic CMB realizations through the same dipole-subtraction, downgrade, and inpainting pipeline and count how many yield 9 or more of the 60 multipoles with Power entropy $p \\le 0.05$; if that fraction is much larger than 0.3%, the reported significance is an artifact of the mask-inpainting procedure rather than intrinsic anisotropy. A more direct check: for a single isotropic simulation, compare Power entropies and principal eigenvectors computed from the full-sky map and from its masked-and-inpainted version; large differences falsify the preservation assumption.","tokens_in":13339,"feed_emoji":"🌌","tokens_out":10575,"duration_ms":100454,"temperature":0.7,"pith_summary":"The paper asks whether the long-claimed \"anomalies\" in the cosmic microwave background's largest angular scales survive in the most recent Planck data release. Applying the Power tensor method to the Planck PR4 Commander temperature map, it finds that nine of the first sixty multipoles ($l=2$ to 61) have anomalously low Power entropy, meaning their power is concentrated along a preferred axis rather than spread isotropically. The chance of finding this many anomalous modes under the hypothesis that the sky is statistically isotropic is about 0.3%, a rejection near the $3\\sigma$ level. The same analysis finds no statistically significant planar modes and no preferred collective alignment axis for the sixty multipoles, though the quadrupole-octopole alignment persists.","feed_headline":"Nine low CMB multipoles break statistical isotropy at 3-sigma","feed_subtitle":"Re-analysis of Planck PR4 finds the old axis and axiality anomalies persist, but no planar-type anisotropy.","key_machinery":"The central object is the Power tensor $A_{ij}(l)$, a $3\\times 3$ matrix built from each multipole's spherical-harmonic coefficients and angular-momentum matrices; it maps a multipole onto an ellipsoid whose normalized eigenvalues define a Power entropy $S(l)=-\\sum_\\alpha \\lambda_\\alpha \\log\\lambda_\\alpha$. Low $S(l)$ means power is concentrated along a single axis (the principal eigenvector, PEV), i.e., the mode is intrinsically anisotropic. The Alignment tensor $X_{ij}=(1/N_l)\\sum_l \\tilde{e}_i^l \\tilde{e}_j^l$, formed by averaging the PEVs of many multipoles, tests for collective preferred directions. The paper also uses a binomial cumulative probability to convert the number of modes falling below a per-multipole $p$-value threshold into an overall significance for the first sixty multipoles.","core_discovery":"The paper reports that in the Planck PR4 Commander-cleaned temperature map, the multipoles $l=6, 13, 14, 16, 17, 30, 34, 40,$ and $56$ have Power entropy $p$-values $\\le 0.05$ against isotropic simulations. Using the binomial distribution with $n=60$ multipoles and a per-multipole threshold $P=0.05$, the cumulative probability of finding $k^*=9$ or more such modes is $P_{\\mathrm{cuml}} \\approx 0.0028$ (Eq. 7), i.e., statistical isotropy is rejected at close to $3\\sigma$. The modes with anomalously small smallest eigenvalues of the Power tensor ($l=13,14,16,30,34,56$) are a subset of these anisotropic modes, so the paper concludes they are axially anisotropic rather than intrinsically planar. The quadrupole and octopole remain aligned at $\\theta_{23} \\approx 14.8^\\circ$ with $p$-values of 0.08 (100 Planck simulations) and 0.035 (5000 ideal realizations), while higher multipoles aligned with the quadrupole are collectively insignificant ($P_{\\mathrm{cuml}} \\approx 0.34$). The Alignment tensor over $l=2$–61 yields no significant preferred axis or plane ($p \\approx 0.21$ and 0.20), and its principal axis lies within about $6^\\circ$–$35^\\circ$ of the anomalous low-multipole axes, broadly close to the dipole, quadrupole, and octopole directions.","pith_inferences":["Extending beyond the paper: the binomial calculation treats the 60 multipoles as independent, but the galactic mask and inpainting introduce correlations between neighboring $l$, so the effective number of independent modes is smaller; a Monte Carlo that directly counts low-entropy modes per simulated map would give a more faithful significance.","Extending beyond the paper: because the same anomaly set was originally identified in WMAP and earlier Planck releases, the threshold and multipole range are not prior-independent; a full look-elsewhere treatment over all exploratory statistics run on these data would reduce the post-trial significance.","Extending beyond the paper: the near-coincidence of the collective alignment axis with the CMB kinetic dipole direction hints at a possible Doppler or aberration origin for the axiality; repeating the Power tensor analysis on dipole-removed maps under different assumptions about the solar velocity would test this.","Extending beyond the paper: applying the same pipeline to CMB polarization, once signal-to-noise allows, would give an independent check of whether the same multipoles are anisotropic in polarization; if the anomalous set differs, residual foregrounds or temperature-map processing are the likely source."],"forward_implications":["Statistical isotropy of the CMB temperature sky is rejected at roughly the $3\\sigma$ level by the Planck PR4 low multipoles, lending support to earlier WMAP-era and Planck 2018 anomaly claims.","The anisotropy is axial rather than planar: every unusually planar mode is also one of the low-entropy anisotropic modes, so explanations that predict planar low-multipole structure are not supported by this data.","The quadrupole-octopole alignment, one of the original CMB anomalies, persists in the newest Planck map at marginal significance, suggesting it is not a processing artifact of earlier releases.","No significant collective alignment axis exists for the full set of sixty multipoles, so the 'axis of evil' picture is not strengthened as a global preferred direction; whatever is happening is concentrated in specific modes.","The collective alignment axis that does emerge lies close to the directions of the dipole, quadrupole, octopole, and the four quadrupole-aligned modes, giving future polarization experiments a concrete axis to test."],"supporting_citations":[{"why":"Introduces the Power tensor, the quadratic estimator that maps each multipole onto an ellipsoid and underlies all the paper's statistics.","marker":"[12]"},{"why":"Develops the Power tensor approach for testing CMB isotropy and supplies the earlier low-multipole results this paper updates.","marker":"[21]"},{"why":"Provides the slightly altered Power tensor and Alignment tensor definitions used in the present analysis.","marker":"[22]"},{"why":"The Planck PR4 (NPIPE) data release that supplies the Commander CMB temperature map and the 100 complementary simulations.","marker":"[20]"},{"why":"The sparse inpainting technique used to fill the masked galactic region; its claim to preserve statistical properties is the paper's key assumption.","marker":"[34]"},{"why":"Planck 2018 cosmological parameters used to generate the 5000 ideal CMB realizations for significance estimates.","marker":"[26]"},{"why":"The previous Planck 2018 assessment of isotropy and statistics that this study re-examines on the newer PR4 map.","marker":"[8]"},{"why":"The original report of low-multipole alignments, providing the quadrupole-octopole axis the paper tests.","marker":"[9]"}],"fun_headline_variants":["CMB low multipoles: 9 modes break isotropy at 3-sigma in PR4","Planck PR4 re-analysis finds axis anomaly still present","Three-sigma anisotropy persists in Planck PR4 temperature map","Quadrupole-octopole alignment robust in Planck PR4 data"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The analysis assumes that the sparse inpainting of the masked galactic region preserves the statistical properties of the low-multipole coefficients, so that Power tensor statistics computed from the inpainted map and from identically processed simulations are unbiased.","fun_headline_variants_meta":{"raw":{"variants":["CMB low multipoles: 9 modes break isotropy at 3-sigma in PR4","Planck PR4 re-analysis finds axis anomaly still present","Three-sigma anisotropy persists in Planck PR4 temperature map","Quadrupole-octopole alignment robust in Planck PR4 data"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000753,"raw_usage":{"total_tokens":3462,"prompt_tokens":1167,"completion_tokens":2295,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":783,"completion_tokens_details":{"reasoning_tokens":2216}},"tokens_in":783,"tokens_out":2295,"duration_ms":18560,"temperature":1.0,"reasoning_tokens":2216,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-06T21:57:47.022884+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Run many statistically isotropic CMB realizations through the same dipole-subtraction, downgrade, and inpainting pipeline and count how many yield 9 or more of the 60 multipoles with Power entropy $p \\le 0.05$; if that fraction is much larger than 0.3%, the reported significance is an artifact of the mask-inpainting procedure rather than intrinsic anisotropy. A more direct check: for a single isotropic simulation, compare Power entropies and principal eigenvectors computed from the full-sky map and from its masked-and-inpainted version; large differences falsify the preservation assumption.","supporting_citations":[{"cited_title":"Ralston and Pankaj Jain","cited_arxiv_id":null,"evidence_quote":"Introduces the Power tensor, the quadratic estimator that maps each multipole onto an ellipsoid and underlies all the paper's statistics."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Develops the Power tensor approach for testing CMB isotropy and supplies the earlier low-multipole results this paper updates."},{"cited_title":"Aluri, John P","cited_arxiv_id":null,"evidence_quote":"Provides the slightly altered Power tensor and Alignment tensor definitions used in the present analysis."},{"cited_title":"Akrami, et al","cited_arxiv_id":null,"evidence_quote":"The Planck PR4 (NPIPE) data release that supplies the Commander CMB temperature map and the 100 complementary simulations."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"The sparse inpainting technique used to fill the masked galactic region; its claim to preserve statistical properties is the paper's key assumption."},{"cited_title":"Aghanim, et al","cited_arxiv_id":null,"evidence_quote":"Planck 2018 cosmological parameters used to generate the 5000 ideal CMB realizations for significance estimates."},{"cited_title":"Akrami, et al","cited_arxiv_id":null,"evidence_quote":"The previous Planck 2018 assessment of isotropy and statistics that this study re-examines on the newer PR4 map."},{"cited_title":"Significance of the largest scale CMB fluctuations in WMAP","cited_arxiv_id":null,"evidence_quote":"The original report of low-multipole alignments, providing the quadrupole-octopole axis the paper tests."}],"review_version":1}