{"id":"339c970a-7246-4b43-b8fb-fd41c33f7a93","arxiv_id":"2607.15564","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":5,"one_line_summary":"25 repeating changing-look AGNs (22 new) show broad-line transitions tracking multi-year optical excursions, with Hβ disappearing at higher accretion rates than Hα.","lead":"Astronomers identified 25 galaxies whose central black holes flip between bright and dim spectral states more than once, 22 of them newly found, by combining archival light curves with fresh spectra. The repeated flips track multi-year brightening and fading of the nucleus and support the 'breathing' picture in which the gas near the black hole crosses a visibility threshold as the feeding rate changes.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Fig. 6 K-S separation may reflect virial-calibration offsets rather than a line-dependent BLR threshold; no per-calibrator breakdown is given.","rationale":"The reader's verdict is already CONDITIONAL, and its weakest-assumption diagnosis—heterogeneous virial calibrations contaminating the K-S test—is exactly the load-bearing concern. I agree with that identification. The rest of the paper is methodologically careful: the Rs>0.3 flux-change threshold, visual inspection of every multi-epoch spectrum, [O III] checks, and the open discussion of the nine non-confirmed candidates all strengthen the sample identification. The seven well-sampled RCL sources showing Eddington-ratio changes tracking photometric excursions are a credible consistency check. However, the broader physical claim about line-dependent BLR visibility thresholds is not established unless the λEdd comparison is robust to calibration and selection effects. The paper gives no per-calibrator breakdown, so the observed D=0.667 could in principle be produced by a 0.3–0.5 dex systematic shift correlated with the Hβ-only vs Hα+Hβ grouping. This does not mean the claim is false—the separation may well survive a uniform-calibration rerun—but it must be tested before the central interpretation is accepted. Because the reader already requested essentially this condition, my verdict is unchanged.","tokens_in":28881,"tokens_out":7394,"duration_ms":88204,"concrete_test":"For every source in Fig. 6, obtain the exact emission line and formula used for MBH and Lbol. Then rerun the K-S and Anderson-Darling tests on: (a) the subset sharing one calibrator and one bolometric correction (e.g., Hα GH05 or on-state Hβ VP06, with L5100); and (b) the full sample after applying published cross-calibration corrections to place all masses on a common scale. Separately, add a 0.4–0.5 dex systematic term in quadrature to every λEdd and recompute D and p. If the Hβ-only vs Hα+Hβ separation persists at p<0.01 in the uniform-calibrator subset and after the systematic floor is added, the line-threshold interpretation is robust; if not, the central Fig. 6 claim is not supported.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The headline statistical claim is the K-S test in §4.2/Fig. 6: off-state log λEdd for Hβ-only vs Hα+Hβ transitions separate at D=0.667, p=2.17e-5. This comparison presupposes that the λEdd values are inter-comparable across the two groups. They are not: §3.1 mixes Vestergaard & Peterson (2006) Hβ, Greene & Ho (2005) Hα, and Vestergaard & Osmer (2009) Mg II virial calibrations, with bolometric corrections 9.26 L5100 and 5.15 L3000. The RCL points fix MBH to the on-state value, while the Dong et al. (2025b) CL points use their home-paper estimates. Single-epoch virial masses have ~0.4 dex intrinsic scatter, and published offsets among these calibrators are of the same order as the group separation implied by D=0.667. If, for example, the Hβ-only group is dominated by Hα-based masses and the Hα+Hβ group by MgII-based masses, a calibration offset could create or inflate the separation with no line-dependent threshold. The paper reports no per-calibrator breakdown, and no redshift or survey-selection control is shown, so it is unclear whether the two groups are otherwise matched. Since the central 'extreme BLR breathing' interpretation rests on this statistical separation, this is the most load-bearing concern.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper searches the Guo et al. (2024, 2025a) and Dong et al. (2025b) changing-look AGN samples for objects whose broad Hβ turned off/on more than once. Using long-baseline ZTF/ATLAS/CRTS optical light curves, WISE/NEOWISE MIR light curves, archival SDSS/LAMOST/DESI spectra, and new LJT/DOT/HCT spectroscopy, the authors select 34 candidates, confirm 25 RCL AGNs (22 new), and classify nine non-confirmed cases. They report that the second transitions occur mostly on rest-frame timescales of 3–4 yr, that seven well-sampled objects show spectral states tracking smooth multi-year optical excursions, and that Hβ-only transitions occur at higher off-state Eddington ratios than Hα+Hβ transitions, based on a K-S test (D=0.667, p=2.17e-5). They interpret this as evidence for a line-dependent BLR visibility threshold and 'extreme BLR breathing', i.e., the BLR gas persists and re-emits rather than being destroyed and rebuilt.","tokens_in":29094,"tokens_out":5822,"duration_ms":71822,"significance":"If the statistical separation in Fig. 6 is robust, the paper would provide the largest sample of newly discovered repeating CL AGNs and a direct, repeated-nucleus test of line-dependent BLR visibility thresholds. The identification pipeline is genuinely careful: Rs>0.3 thresholding, visual inspection of all multi-epoch spectra, [O III] checks against data-quality artifacts, and individual discussion of the nine non-confirmed candidates are all strengths. The seven well-sampled sources with continuous light curves offer a useful link between spectroscopic state and continuum excursion. However, the headline K-S result compares Eddington ratios that are not placed on a common virial-calibration scale, and the paper does not show that the separation survives homogeneous recalibration. The central interpretive claim is therefore currently not established, though it is plausibly fixable with additional analysis.","major_comments":[{"comment":"The Anderson-Darling p-value is reported as '≤ 10^-3 (floored by numerical resolution)', which is inadequate for a claim of 'confidence level exceeding 99.9%'. A floor of 10^-3 is only 99.9% if the true value is exactly 0.001; reporting a one-sided inequality without a better determination, or using a Monte Carlo estimate, is not sufficient to support the stated confidence level. This is a secondary issue relative to the calibration mixing, but it should be corrected in any revision.","section":"§4.2 / Fig. 6"}],"minor_comments":[{"comment":"'As Seven RCL AGNs' should read 'Seven RCL AGNs' (capital A and misplaced 'As').","section":"Abstract"},{"comment":"Header 'Seciont' is a typo for 'Section'.","section":"§2"},{"comment":"'solide curves' should be 'solid curves'.","section":"Fig. 5 caption"},{"comment":"The solar-mass symbol appears as 'uni2299' in several axis labels and table headers; these are unicode rendering artifacts that should be fixed.","section":"General"},{"comment":"The definition of Rs as 'Sb − Sd / Sb' is ambiguous; it should be written as (Sb−Sd)/Sb.","section":"Eq. (1)"},{"comment":"Several entries contain stray spaces ('5953 2', '6111 9a', '5739 5a'); the table needs a formatting pass.","section":"Table A1"},{"comment":"'observed (rest-frame) timescales' mixes observer-frame and rest-frame terminology; the sentence should specify which quantity is which.","section":"§3.2"},{"comment":"The statement that derived data are available 'on reasonable request' is weaker than the reproducibility standards common in this field; public tables or a machine-readable supplement would be preferable.","section":"Data Availability"}],"recommendation":"major_revision","confidential_remarks":"The discovery sample and the careful identification work are solid, and I would not reject the paper. However, the headline physical claim about line-dependent BLR visibility thresholds currently rests on a K-S test performed on Eddington ratios that are not on a common calibration scale. This is fixable with a per-calibrator breakdown, a homogeneous recalibration, or an explicit demonstration that the separation survives such corrections. If the authors can provide that, the paper would be a valuable contribution. The circularity concern in Fig. 3 should also be addressed by rephrasing or by an independent mass/continuum test."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The useful thing here is the sample: 25 confirmed repeating changing-look AGNs, 22 new, found by systematic light-curve selection from known CL samples and confirmed with new spectra. The identification is careful—quantitative Rs threshold, visual inspection, [O III] checks, and the nine rejected candidates are discussed individually. That is a real contribution; anyone working on CL AGNs will want this catalog.\n\nThe BLR breathing interpretation is plausible and consistent with Elitzur & Ho and the Guo et al. sequence, and the seven well-sampled sources showing spectral state tracking the optical continuum are a nice piece of evidence.\n\nThe main weak spot is the K-S test in Fig. 6. The off-state λEdd values mix Hβ, Hα, and Mg II virial calibrations with different bolometric corrections, and the RCL points fix MBH to the on-state value while the CL comparison points use their home-paper estimates. Systematic offsets among those calibrators are of the same order as the reported group separation. The paper gives no per-calibrator breakdown, so it is unclear whether the Hβ-only vs Hα+Hβ separation is a line-dependent threshold or a calibration artifact. That is load-bearing for the abstract's claim. It needs a restricted test on a uniformly calibrated subsample, or at least a demonstration that the separation survives within calibrator.\n\nThe “mostly 3–4 yr” intervals in the abstract are largely a cadence artifact; the photometric dt1 estimates are the more physical timescale and the paper knows this. The data availability statement should be improved; confirming spectra and derived catalog should go to an archive.\n\nMinor: the Fig. 3 consistency check is partly built into the measurement—on/off defined by Hβ visibility tracks the same continuum that sets λEdd—so I wouldn't lean on it as independent confirmation, but the paper doesn't lean on it heavily.\n\nThis deserves a serious referee. The sample is valuable and the identification work is solid; the statistical claim needs to be tested against calibration systematics before publication. I'd send it out.","headline":"A useful new RCL sample with careful follow-up, but the headline Eddington-ratio separation rests on a virial-calibration comparison that needs a robustness check.","tokens_in":29818,"tokens_out":2741,"would_cite":true,"duration_ms":28737,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"Repeating changing-look AGNs show broad-line regions that persist and breathe rather than being destroyed and rebuilt","keywords":["changing-look AGN","repeating changing-look AGN","broad-line region","BLR breathing","Eddington ratio","AGN variability","quasar spectroscopy","supermassive black holes"],"falsifier":"Recompute the Figure 6 comparison with one uniform virial calibration and one bolometric correction applied to all 74 sources; if the Hβ-only and Hα+Hβ off-state Eddington-ratio distributions then overlap (K-S p > 0.05), the line-dependent visibility threshold is not established. A complementary check: for the seven continuously monitored RCL AGNs, predict a third transition at the next optical light-curve extremum and observe whether the broad lines reappear or disappear with the continuum; if they do not, the breathing picture is falsified.","tokens_in":28617,"feed_emoji":"🔭","tokens_out":5267,"duration_ms":57302,"temperature":0.7,"pith_summary":"This paper identifies 25 repeating changing-look AGNs, 22 of them new, in which the same nucleus turns its broad emission lines off and on again, some within three to four years and possibly in as little as a few months. Its central claim is that these repeated transitions are not destruction and rebirth of the broad-line region (BLR), but 'breathing': the same BLR gas persists while the central accretion power rises and falls, carrying individual emission lines across visibility thresholds. The key quantitative result is a statistical separation between Hβ-only transitions and transitions involving both Hα and Hβ: Hβ-only events happen at higher off-state Eddington ratios, so Hβ fades from view at a higher accretion level than Hα. Seven sources with nearly continuous light curves show their spectral on/off states tracking multi-year optical excursions, with infrared dust emission following the same trends. If true, changing-look AGNs become a window onto a reversible, line-dependent BLR response to accretion variability rather than a catastrophic reconfiguration of the gas.","feed_headline":"AGN broad lines breathe: 25 repeating changers, 22 new","feed_subtitle":"Same nuclei flicker on and off as H-beta fades before H-alpha, tracing a line-dependent visibility threshold","key_machinery":"The load-bearing instrument is the RCL AGN sample itself: because the same nucleus crosses the same physical boundary more than once, one-off destructive events or simple line-of-sight obscuration are excluded as explanations. Within that sample, the discriminant is the two-sample K-S test on off-state log Eddington ratios (D = 0.667, p = 2.17 × 10⁻⁵) between Hβ-only and Hα+Hβ transition groups, supplemented by an Anderson-Darling test. The proposed physical mechanism is BLR breathing: a radially stratified broad-line region whose line emissivities, responsivities, and detectability vary with the incident ionizing flux, so lines go below visibility thresholds in a fixed order as accretion fa","core_discovery":"The paper establishes, on its own terms, that repeated changing-look transitions — on–off–on or off–on–off sequences in the same active galactic nucleus — trace reversible changes in the central ionizing continuum. The broad-line region is not removed and re-formed each time; instead, the same structure 'breathes': as the ionizing flux weakens, broad Hβ falls below its detection threshold first, at a relatively high Eddington ratio, while broad Hα remains visible until the source sinks to a lower accretion level. The authors demonstrate this with a two-sample Kolmogorov-Smirnov test on off-state Eddington ratios (D = 0.667, p = 2.17 × 10⁻⁵) and an Anderson-Darling test, comparing Hβ-only and","pith_inferences":["A natural next test, not run in the paper, is whether Hβ-only, Mg II-only, and Hα-only transitions form an ordered ladder in off-state Eddington ratio; the paper's line-responsivity argument predicts Mg II sits between Hβ and Hα.","The K-S comparison mixes three black-hole-mass calibrations and two bolometric corrections; recomputing Figure 6 with a single uniform calibration, or separately by group, would settle whether the claimed threshold separation survives calibration systematics.","If BLR breathing is correct, high-cadence reverberation mapping across a second transition should show the emission-line lag, width, and responsivity returning toward their first-epoch values as the continuum recovers — a test that distinguishes breathing from stochastic gas rearrangement.","The same-nucleus logic could be extended to predict future transitions: unconfirmed candidates with moderate continuum changes are natural targets for monitoring at the next light-curve extremum.",""],"forward_implications":["A single changing-look event cannot distinguish destroyed gas from faded line emission; repeated transitions directly demonstrate that the broad-line region persists, so CL events should be treated as accretion-state diagnostics rather than evidence of BLR destruction.","Because Hβ fades at a higher Eddington ratio than Hα, off-state classification depends on which lines a spectrum covers; surveys that only measure Hβ will classify sources as 'off' earlier than those including Hα, so wavelength coverage must be reported in CL-AGN catalogs.","The driver of CL transitions must be reversible and operate on month-to-year timescales; candidate processes such as accretion-mode changes, heating/cooling fronts, magnetically supported disks, or radiation-pressure instability are constrained to produce smooth multi-year excursions rather than stochastic flickering.","The latest spectra of RCL AGNs fall on the expected side of the CL population in the Eddington-ratio versus black-hole-mass plane, supporting the use of optical light-curve extrema as predictors of spectroscopic state.","The line-dependent threshold sequence implies that Hβ, and later Hα, can be used as accretion-level indicators: when Hβ disappears while Hα remains, the source sits in an intermediate accretion state.",""],"fun_headline_variants":["BLR breathing: 22 new repeat changing-look AGNs","Same AGN toggles on/off: 22 new repeaters found","Hβ fades first: 22 new repeating AGN reveal BLR breathing","Repeating AGN flickers: line-dependent threshold found","AGN on-off cycles: 22 new cases, BLR breathes"],"cache_read_input_tokens":2304,"weakest_assumption_plain":"The load-bearing premise is that the two Eddington-ratio distributions compared in Figure 6 are measured on a common scale: the groups mix three black-hole-mass calibrations and two bolometric corrections, with RCL masses pinned to on-state values, so a systematic shift of roughly 0.3–0.5 dex correlated with the Hβ-only versus Hα+Hβ grouping would erase the threshold signal.","fun_headline_variants_meta":{"raw":{"variants":["BLR breathing: 22 new repeat changing-look AGNs","Same AGN toggles on/off: 22 new repeaters found","Hβ fades first: 22 new repeating AGN reveal BLR breathing","Repeating AGN flickers: line-dependent threshold found","AGN on-off cycles: 22 new cases, BLR breathes"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000463,"raw_usage":{"total_tokens":2210,"prompt_tokens":865,"completion_tokens":1345,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":609,"completion_tokens_details":{"reasoning_tokens":1248}},"tokens_in":609,"tokens_out":1345,"duration_ms":10713,"temperature":1.0,"reasoning_tokens":1248,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-01T22:58:18.980427+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Recompute the Figure 6 comparison with one uniform virial calibration and one bolometric correction applied to all 74 sources; if the Hβ-only and Hα+Hβ off-state Eddington-ratio distributions then overlap (K-S p > 0.05), the line-dependent visibility threshold is not established. A complementary check: for the seven continuously monitored RCL AGNs, predict a third transition at the next optical light-curve extremum and observe whether the broad lines reappear or disappear with the continuum; if they do not, the breathing picture is falsified.","supporting_citations":[],"review_version":1}