{"id":"48c23bc9-1040-4f4a-8ac7-6585f89948c2","arxiv_id":"2506.06415","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":5,"one_line_summary":"Kepler red giants include stars whose envelopes rotate faster than their cores, plus clump stars with unusually fast-spinning cores, both of which strain standard angular momentum transport models.","lead":"Using a neural network trained on synthetic spectra, the authors measure core and envelope rotation in 1,517 Kepler red giants. They report two unusual groups: clump stars with very fast-spinning cores, and giants whose outer envelopes rotate faster than their cores, which challenges current models of angular momentum transport.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Network accuracy is not established in the anomalous regime; the population-level anomaly claim rests on unvalidated extrapolation.","rationale":"The individual anomalous stars (KIC 11615944, KIC 11546972, KIC 6695665) are corroborated by MCMC fits and visual inspection of the splittings, so the existence claim for a few stars is reasonably supported. However, the MCMC and the network share the same asymptotic two-zone generative model and the same priors; MCMC agreement therefore does not independently validate the model for these stars. The population-level statement--that there is 'a group of clump stars' with rapid cores and that 'some' giants show inverted ratios--requires the network to be unbiased in the anomalous parameter region. The paper does not demonstrate this: synthetic validation is aggregated over the training distribution, and the specific stress test in Appendix D uses a high-Delta_nu RGB-like case. The most serious failure mode is the splitting/spacing degeneracy described in Section 3.1, which can alias a high-inclination, slowly rotating star to a lower-inclination, faster rotating solution. With no external catalog overlap for the anomalies, the network's confident outputs are the sole basis for the population claim. This is an addressable concern: a dedicated synthetic test in the anomalous regime, as outlined above, would settle whether the network's calibration holds exactly where it matters. If the test passes, the claim is substantially strengthened; if it fails, the anomalies should be reinterpreted as calibration artifacts rather than astrophysical discoveries. The reader's CONDITIONAL verdict is appropriate; I would not change it without the test. Credit is due for the transparent limitation statements, the MCMC checks, and the synthetic noise tests, but these do not cover the critical regime.","tokens_in":110599,"tokens_out":8917,"duration_ms":89744,"concrete_test":"Build an anomaly-focused synthetic test set: draw 2000 spectra with Delta_nu in [1,8] microHz, Delta_Pi in [150,500] s, q in [0.1,0.6], Omega_core/2pi in [0.005,2.8] microHz, Omega_env/2pi in [0.005,0.4] microHz (with ~30% of cases Omega_env>Omega_core and ~20% Omega_core>0.5 microHz), noise and observation times sampled from Kepler red giants. Run the already-trained network and compare inferred vs injected Omega_core and Omega_env. Report bias, RMS scatter, and the false-positive rate for Omega_env>Omega_core. If the inferred values track the injected values within the stated 1-sigma uncertainties in this region, the concern is resolved; if the network preferentially produces inverted ratios for slow cores, or underestimates high core rotation, the anomaly population is likely spurious.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim requires the CNN to be accurate precisely for the unusual stars that define the anomalies: low-Delta_nu clump stars with Omega_env > Omega_core, and clump stars with Omega_core/2pi near 1 microHz. Section 2.2 states the network 'performs reliably only for stars that fall within this simulated parameter space' and that 'unexpected internal rotation' may produce biased predictions. The anomalous rotators are exactly such cases, and Figure 33 shows they have zero overlap with the Gehan et al. (2018) catalog, so no external validation exists. The network's synthetic validation (Figure 2 and Appendix D) is averaged over the full training distribution and is not conditioned on the anomalous region; the stress test in Appendix D uses Delta_nu = 16 microHz and Delta_Pi = 85 s, i.e., the RGB regime, not the clump regime. Moreover, Section 3.1 describes a degeneracy between rotational splitting and mixed-mode spacing that can make a spectrum at one inclination look like a different rotation rate at another. This degeneracy, unresolved by external comparison for the anomalous stars, could systematically map normal stars into the anomalous region. Since the population-level claims (systematic ratio changes, a group of fast clump cores) rest on these network outputs, the claim is unsupported until the network is validated where the anomalies live.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper develops a convolutional neural network that infers core and envelope rotation rates, along with other seismic parameters, directly from Kepler power spectra of red giants. The network is trained on ~5 million synthetic spectra built from asymptotic theory, applied to ~21,000 Kepler targets, and a 'confident' subset of 1,517 stars is analyzed. The authors report that most measurements reproduce existing core-rotation catalogs, but they also identify two claimed anomaly populations: red giants and clump stars with envelope rotation faster than core rotation, and clump stars with rapidly rotating cores (Ω_core/2π near 1 μHz). They verify three individual cases with MCMC fits and visual splitting inspection, and they propose weak angular-momentum transport or binary interactions as explanations, supported by MESA single-star and binary models. The central claim is that these anomalies challenge standard angular-momentum transport theory and have implications for compact remnant spins.","tokens_in":110832,"tokens_out":2552,"duration_ms":27907,"significance":"If the anomalies are real, the paper would be significant: it would extend the established two-zone rotation picture for red giants (Ω_env < Ω_core in essentially all published stars) to include inverted rotation profiles and clump cores rotating 10-20 times faster than the median, with consequences for angular-momentum transport prescriptions and the spins of white dwarfs and neutron stars. The paper has several concrete strengths: the training simulator is available, the MCMC follow-up of three anomalous stars is a good-faith attempt at external verification, the visual inspection of splittings adds transparency, and the comparison with the Gehan et al. (2018) catalog provides a quantitative benchmark for the bulk of the sample. These strengths make the paper a useful contribution even if the anomaly population requires further validation. However, the population-level claims rest entirely on the reliability of the CNN in a regime where the authors themselves state it may be biased, and the external validation does not cover that regime; therefore the significance is conditional on an unresolved validation gap.","major_comments":[{"comment":"The load-bearing population-level claims require the CNN to be accurate for the anomalous stars, but the paper itself states that the network performs reliably only within the simulated parameter space and that 'unexpected internal rotation' may yield biased predictions. The anomalous rotators are precisely the 'unexpected internal rotation' cases, and Figure 33 shows that none of the envelope-super-rotation stars overlap with the Gehan et al. (2018) catalog, so there is no external validation for them. The synthetic validation in Figure 2 and Appendix D is averaged over the training distribution, not conditioned on the anomalous region, and the stress test in Appendix D uses Δν = 16 μHz, ΔΠ = 85 s, which is the RGB regime rather than the clump regime. The authors should validate the network on synthetic spectra drawn from the anomalous parameter space (low Δν, ΔΠ > 150 s, Ω_env > Ω_core, Ω_core/2π near 1 μHz) and report the accuracy and calibration in that region; without this, the anomaly population may be an artifact of extrapolation.","section":"Section 2.2"},{"comment":"The degeneracy between rotational splitting and mixed-mode spacing, described in Section 3.1, is a concrete mechanism by which normal stars could be mapped into the anomalous region. The authors note that a spectrum with splitting equal to one-quarter of the mixed-mode spacing at i=90° can look like one-third splitting at i=55°, and they use this to explain discrepancies with Gehan et al. (2018). However, the same degeneracy could affect the anomalous stars, whose splittings are large relative to the mixed-mode spacing, and for which the network is trained on synthetic mode patterns that may not cover all inclination-splitting combinations. The MCMC fits to three stars do not resolve this at the population level, since MCMC uses the same asymptotic model and the same priors. I request a dedicated synthetic test that quantifies how often normal rotation configurations are misclassified into the anomalous region as a function of SNR, inclination, and mixed-mode spacing.","section":"Section 3.1"},{"comment":"The validation against Gehan et al. (2018) shows that only 59.2% of the 426 overlapping confident stars agree within 20% of the 1:1 relation, while 25.3% fall in no defined proximity zone. This is a substantial disagreement rate, and the paper attributes most discrepancies to the network's use of inclination-dependent amplitudes. Yet this explanation is not demonstrated for the 25.3% category. More importantly, the anomalous stars are not present in this overlap sample at all (Figure 33), so the external validation does not establish reliability for the anomaly population. The authors should show whether the anomalous candidates, when they do have any published counterpart or when re-analyzed with an independent method (e.g., a classical peak-bagging fit), still retain Ω_env > Ω_core and fast clump cores.","section":"Section 3.1"},{"comment":"The selection criteria (pmax thresholds and i > 45°) and the bin sizes for the ordinal classification are presented as choices, but there is no sensitivity analysis. For instance, the minimum pmax(Ω_env/2π) > 0.2 is a low bar for a network that outputs probabilities over bins, and the inclination cut at 45° may interact with the degeneracy noted in Section 3.1. I request that the authors demonstrate that the anomalous population is stable under reasonable variations of these thresholds and bin sizes, e.g., by showing the number of anomalous stars as a function of the pmax thresholds. Without such a test, it is unclear whether the anomaly population is robust or a threshold artifact.","section":"Section 2.2"}],"minor_comments":[{"comment":"The phrase 'anomalously fast' is used for both envelope-super-rotation and fast clump cores, but the two phenomena are physically distinct; using separate terms (e.g., 'inverted rotation ratio' and 'fast clump cores') would improve clarity.","section":"Abstract"},{"comment":"The sentence 'The network’s distributions, depicted in gray, were juxtaposed against the posterior distributions obtained from MCMC in red' in Figure 1 uses 'gray' and 'red' but the figure is reproduced in black and white in the arXiv version; please ensure the figure caption and text are consistent with the actual rendering.","section":"Section 2.2"},{"comment":"The training range for Ω_core/2π is given as 0.005-2.8 μHz and for Ω_env/2π as 0.005-0.4 μHz. The paper should state whether the training distribution is uniform in these ranges and whether any samples were generated with Ω_env > Ω_core, since the anomaly population requires that configuration.","section":"Section 2.1"},{"comment":"The computation of the rotation-rate ratio assumes independent distributions for Ω_core and Ω_env to obtain the largest uncertainty interval, as shown in Figure 27. This is a conservative choice, but the paper should note that the actual ratio distribution and the probability P(Ω_env/Ω_core > 1) may be different if a physical correlation exists.","section":"Appendix C"},{"comment":"Several entries in Table 3 have very large asymmetric uncertainties (e.g., KIC 10219075 with Ω_core/2π = 0.11 +1.88 -0.07 μHz), and the dagger flag for Nyquist proximity is helpful but only applied to a few rows. The paper should consider flagging or excluding stars with relative uncertainties exceeding a threshold, or at least clearly list them separately.","section":"Table 3"}],"recommendation":"major_revision","confidential_remarks":"The paper is ambitious and presents a large new catalog plus a bold physical claim. The main risk is that the anomaly population is an artifact of the CNN's extrapolation beyond its validated regime, which the authors themselves acknowledge in Section 2.2. The paper would be substantially strengthened by adding a synthetic validation specifically in the anomalous parameter space and by demonstrating that the anomaly population is robust to choices of selection thresholds. I do not think outright rejection is warranted, because the MCMC follow-up of three stars and the visual inspection are credible individual cases, and the authors have been transparent about the network's limitations. However, as it stands, the population-level claim is not yet supported. I would also note that the paper's framing 'anomalously fast' may overstate the case given the 25-40% disagreement rate with the existing catalog in the overlap sample; a more cautious interpretation would help fit the evidence."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Bottom line: the paper is a real attempt at a hard measurement and the largest matched core/envelope sample I know of, but the discovery population rests on network outputs that are not validated where the anomalies live. The individual showcase stars (KIC 11615944, 11546972, 6695665) are the strongest part: MCMC fits and visual splitting checks support envelope-super-rotation and fast-rotating core in those specific objects. That gives the existence claim real weight. What is genuinely new is the pipeline itself—a CNN that maps power spectra directly to rotation parameters—and the catalog of 1,517 stars with both core and envelope rotation. The population-level claims (systematic ratio changes, a distinct class of fast clump cores) are absent from earlier catalogs, so if they hold they matter for angular-momentum transport and compact remnant spins.\n\nThe soft spot is exactly where the stress test points. Section 2.2 says the network is reliable only inside the simulated parameter space, and that unexpected internal rotation may bias predictions. The anomalous stars are unexpected internal rotation by definition, and none of them overlap with Gehan et al. (2018), so there is no external check on the network in that regime. Figure 33 is honest about that, which I appreciate, but it does not fix the problem. The Appendix D stress test uses Delta_nu = 16 microHz and Delta_Pi = 85 s—RGB regime, not the clump regime where the anomalies sit. And the rotation-splitting/mixed-mode-spacing degeneracy described in Section 3.1 means a normal star at one inclination can look like a different rotation rate at another; without external comparison for the anomaly candidates, that degeneracy could push normal stars into the anomalous region. The uncertainty calibration is also trained and tested on the same simulator family, so the quoted error bars on the anomaly population are partly self-referential.\n\nOn the other side: the paper is unusually transparent about its limitations, the MCMC work on the three flagship stars is real, and the partial agreement with Gehan (59% of 426 overlapping stars within 20%) is roughly what one expects from independent pipelines. The problems are addressable: validate the network on clump stars with known rotation, or at least rerun the stress test in the clump regime, and do targeted MCMC or mode-fitting on a larger subsample of the anomalous candidates. Until then I read the fast-clump-core population and systematic ratio trends as interesting but unproven.\n\nThis deserves a serious referee—it is important, methodologically novel, and honest—but it needs heavy revision before publication. I would send it out with a request for clump-regime validation and external anchoring of the anomaly sample.","headline":"A genuine large-sample measurement with transparent caveats, but the anomalous-rotator population is not yet validated where it lives.","tokens_in":111448,"tokens_out":2352,"would_cite":true,"duration_ms":24634,"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":"Asteroseismic survey finds red giants whose envelopes rotate faster than their cores.","keywords":["red giants","red clump stars","asteroseismology","stellar rotation","core-envelope differential rotation","angular momentum transport","neural network","binary interactions"],"falsifier":"Apply the paper's own MCMC forward-fitting, without the neural network, to every claimed anomalous rotator and check whether the $\\ell=2$ p-mode splittings remain larger than the core-dominated $\\ell=1$ splittings; if most inverted-ratio cases disappear or shift below one, the anomaly population is an artifact of the machine-learning prior.","tokens_in":110338,"feed_emoji":"🔄","tokens_out":5589,"duration_ms":59007,"temperature":0.7,"pith_summary":"The paper seeks to establish that the standard two-zone picture of red-giant rotation, in which cores spin much faster than envelopes, is not universal. From space-based photometry of about 1,517 red giants, it reports a population of red-giant and red-clump stars whose envelopes appear to rotate faster than their cores, and a separate group of clump stars whose cores rotate 10 to 20 times faster than the typical clump core. If correct, these anomalies show that angular-momentum transport between core and envelope is not a single universal process, and that some stars either transport angular momentum very weakly or have been spun up by binary interactions. The authors favor weak internal transport for the rapid-core population, noting that their binary models reproduce the fast cores only at the price of over-fast envelopes.","feed_headline":"Red giants found spinning envelopes faster than cores","feed_subtitle":"Survey of 1,517 stars also finds clump cores spinning 10–20 times faster than normal.","key_machinery":"The load-bearing object is the two-zone rotational splitting relation $\\delta\\nu_{\\rm rot} = \\frac{1}{2}\\big(\\frac{\\Omega_{\\rm core}}{2\\pi}\\zeta(\\nu) + \\frac{\\Omega_{\\rm env}}{2\\pi}[1-\\zeta(\\nu)]\\big)$, where the mixing fraction $\\zeta(\\nu)$ interpolates between pressure-dominated envelope modes and gravity-dominated core modes. Envelope rates come mainly from $\\ell=2$ p-dominated splittings and core rates from $\\ell=1$ mixed-mode splittings. On top of that sits a convolutional-LSTM network trained as an ordinal classifier on five million synthetic spectra, which outputs probability distributions for the core rotation, envelope rotation, and four seismic parameters; MCMC fits verify the network's inferences for the anomalous stars, and the rotation-ratio uncertainties are computed from the two marginal distributions under a conservative negative-correlation assumption.","core_discovery":"The central claim is that exceptions to the seismic rotation ordering of red giants are real and form a distinct population. Using a neural network trained on synthetic oscillation spectra and confirmed with MCMC fits for representative stars, the paper infers core and envelope rotation for 1,517 red giants and finds a systematic evolution of the envelope-to-core rotation ratio: it declines along the red-giant branch and then rises to values near 0.01 to 4 in the clump phase. Within that spread sit stars with the envelope-to-core ratio above 1 and clump stars with core rotation near one microhertz, that is, 10 to 20 times the median clump core rate. The paper argues these anomalies challenge current angular-momentum-transport models, that a weaker magnetic-transport prescription can explain the fast cores, and that binary spin-up offers an alternative that is in tension with the observed slow envelopes.","pith_inferences":["A sharper test of the neural-network interpretation would be to retrain it on synthetic spectra with the prior support for $\\Omega_{\\rm env}>\\Omega_{\\rm core}$ removed and check whether the anomalous population still emerges; without such a test, part of the signal could in principle be a regression-to-prior artifact.","The paper computes ratio uncertainties from independent core and envelope marginals, but the joint posterior is likely correlated, so the reported probabilities $P(\\Omega_{\\rm env}/\\Omega_{\\rm core}>1)$ may overstate or understate confidence for individual stars.","Applying the same network to TESS and PLATO red giants would provide a population-level check of how the incidence of envelope-super-rotation varies with evolutionary state, independent of any single Kepler spectrum.","The anomalous-rotator list could be cross-matched with asteroseismic catalogs of subgiants and early red-giant-branch stars to see whether the inverted ratio is acquired at a specific evolutionary transition or inherited from main-sequence binaries."],"forward_implications":["If the anomalous rotators are real, the usual assumption that red-giant cores always rotate faster than their envelopes must give way to a picture with at least two rotation channels: normal strong-transport stars and anomalous weak-transport or spun-up stars.","The fast clump cores, near $\\Omega_{\\rm core}/2\\pi\\sim1\\,\\mu$Hz, lie close to original magnetic-dynamo predictions and about ten times above enhanced-transport models, implying that some stars avoid efficient internal angular-momentum transport.","Binary tidal spin-up and merger models reproduce fast cores but predict envelopes rotating far faster than observed, so the data favor weak transport over binarity for the rapid-core population.","Anomalous rotators show no unusual lithium, carbon-to-nitrogen, or binarity indicators, so binarity does not obviously identify them, and larger samples will be needed to detect any weak abundance or activity correlation.","If the fast cores persist into later evolutionary phases, they would naturally produce rapidly rotating white dwarfs and, in more massive stars, could yield energetic supernovae and gamma-ray bursts."],"supporting_citations":[{"why":"Supplies the published core-rotation catalog used to validate the network's measurements and to identify the parameter regimes where earlier methods disagree.","marker":"Gehan et al. 2018"},{"why":"Provides the stellar-evolution models with enhanced magnetic angular-momentum transport that reproduce normal red-giant and clump rotation rates and define the factor-of-ten discrepancy for the fast rotators.","marker":"Fuller et al. 2019"},{"why":"Establishes the foundational asteroseismic measurement of red-giant core rotation that this work extends to a larger sample.","marker":"Mosser et al. 2012"},{"why":"Supplies the simulator that generates the synthetic power spectra used to train the neural network, making it load-bearing for the method.","marker":"Benomar 2023"},{"why":"Provides the magnetically active stars used to check that strong internal magnetic fields do not systematically bias the network's core-rotation estimates.","marker":"Li et al. 2023"},{"why":"Offers independent measurements of the global seismic parameters used to confirm that the anomalous rotators have otherwise ordinary structure.","marker":"Vrard et al. 2016"},{"why":"Supplies the simulation result that magnetic-dynamo action can bifurcate, supporting the idea that some stars host weak dynamos and therefore rapidly rotating cores.","marker":"Barrère et al. 2023"}],"fun_headline_variants":["Red giant survey finds stars with envelopes outspinning cores","Clump stars with cores spinning 10-20 times faster than normal","Envelope-to-core rotation reverses in some red giants","Red giants with core rotation exceeding envelope rotation"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The neural network is trusted to be unbiased precisely for the unusual stars it was not validated against: the synthetic training data cover normal rotation profiles, and the paper itself warns that stars with unexpected internal rotation may receive biased or inaccurate predictions.","fun_headline_variants_meta":{"raw":{"variants":["Red giant survey finds stars with envelopes outspinning cores","Clump stars with cores spinning 10-20 times faster than normal","Envelope-to-core rotation reverses in some red giants","Red giants with core rotation exceeding envelope rotation"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000705,"raw_usage":{"total_tokens":3121,"prompt_tokens":831,"completion_tokens":2290,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":447,"completion_tokens_details":{"reasoning_tokens":2224}},"tokens_in":447,"tokens_out":2290,"duration_ms":16644,"temperature":1.0,"reasoning_tokens":2224,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-07T05:58:28.188422+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Apply the paper's own MCMC forward-fitting, without the neural network, to every claimed anomalous rotator and check whether the $\\ell=2$ p-mode splittings remain larger than the core-dominated $\\ell=1$ splittings; if most inverted-ratio cases disappear or shift below one, the anomaly population is an artifact of the machine-learning prior.","supporting_citations":[],"review_version":1}