{"id":"afafe1bb-fdd7-47d0-88c5-a0f98be2fe53","arxiv_id":"2509.03571","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":4,"one_line_summary":"Super-Eddington accretion models can explain why little red dots show almost no variability, whereas standard sub-Eddington AGN variability models predict changes that should already have been seen.","lead":"This paper asks why JWST's 'little red dots' rarely show brightness changes, even though they look like black-hole-powered galaxies. The authors simulate light curves and find that super-Eddington black hole accretion naturally produces the observed quietness, with a testable prediction for upcoming JWST surveys.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Super-Eddington optical PSD rests on a two-decade radial extrapolation (10–200 to 10^4 r_g); if the true PSD at 10^4 r_g is only moderately steeper or higher-normalized, the predicted non-detection fractions no longer hold.","rationale":"The reader's weakest_assumption identifies the same load-bearing issue: the super-Eddington optical variability model is obtained by a large radial extrapolation of PSD parameters from 10–200 r_g to 10^4 r_g, with no direct simulation or observation in the optical region. I concur that this is the weakest link in the central claim. The paper's own Section 4 caveat confirms the assumption but only explores an alternative that makes variability even smaller, leaving the opposite error mode untested. This is a genuine correctness risk, not merely a disagreement with consensus: if the true optical PSD were steeper in normalization or shallower in slope, the predicted detection fractions for KH24 and NEXUS could rise above the observational thresholds. The DRW comparison and the transparent treatment of caveats are strengths, and the qualitative conclusion about sub-Eddington DRW models is supported. The verdict should remain CONDITIONAL: the model is plausible but requires validation of the PSD extrapolation before the super-Eddington interpretation is preferred. I therefore do not change the reader's verdict.","tokens_in":21778,"tokens_out":5782,"duration_ms":68822,"concrete_test":"Run a global radiation-MHD disk simulation with the Jiang et al. (2025) setup but with the emitting region extending to ~10^4 r_g at Mdot≈3 M_Edd, and directly fit the optical PSD, PSD∝C0 ν^-β, over rest-frame frequencies ~1 day^-1 to 1 yr^-1. Then recompute the KH24 and NEXUS Δm distributions of §3.2.1 using the measured β and C0 instead of the extrapolated values. If the directly measured PSD yields more than ~1% of KH24 mock light curves with Δm>0.21 mag (current super-Eddington value is <1%) or more than ~12% of NEXUS light curves with Δm>0.21 mag, the match to non-detections is an artifact of the extrapolation and the central claim is not established. A cheaper complementary check is to compare the extrapolated optical PSD against rest-frame optical structure functions of local super-Eddington AGN (e.g., narrow-line Seyfert 1s with L/L_Edd≈1–10) over 0.1–10 yr timescales.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim that super-Eddington accretion can explain the lack of LRD variability depends on the optical PSD constructed in §2.2.2. The no-variability results (⟨Δm⟩=0.062 and <1% >3σ for KH24; 12% for NEXUS) are controlled by the extrapolated PSD ∝ 5×10^-9 ν^-3 at 10^4 r_g. This is obtained by logarithmically extrapolating β and C0 from simulation PSDs at <10 r_g (soft X-ray) and 50–200 r_g (UV), spanning roughly two decades in radius beyond the outermost anchor. The two anchors already differ in both slope (β=2.4→2.6) and normalization (C0=10^-6→10^-7), so the extrapolation has no empirical or simulational anchor in the optical region. Section 4 acknowledges the assumption but tests only an alternative that makes variability even smaller (constant amplitude, τ∝r^3/2). That does not bound the opposite direction: if the optical annulus has a steeper/warmer fluctuation spectrum, residual irradiation, or wind-driven variability, C0 or β could be larger, pushing Δm above the 3σ threshold for a substantial fraction of mock light curves. The DRW comparison (33–47% predicted detections vs none observed) independently disfavors the sub-Eddington DRW model, but it does not by itself validate the super-Eddington model. Thus the unvalidated radial extrapolation is the load-bearing assumption for the paper's positive claim.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper investigates whether the observed lack of variability in JWST-discovered little red dots (LRDs) can be explained by super-Eddington accretion. The authors generate mock light curves using two models: (i) a Damped Random Walk (DRW) model calibrated to lower-redshift sub-Eddington AGN (Burke et al. 2021, 2023) and (ii) a novel super-Eddington model in which the optical PSD is constructed by extrapolating radiation-MHD simulation results from the inner disk (Jiang et al. 2019, 2025). Comparing with existing observations (T24, KH24), they find that the DRW model predicts detectable variability in 33-47% of mock light curves, while none is observed; the super-Eddington model predicts <1% detections for current KH24-like cadences and 12% for the planned NEXUS survey. They also predict that the TWINKLE campaign will detect broad-line variability if a soft X-ray driver reaches the broad-line region. The paper concludes that super-Eddington accretion can naturally account for the lack of observed continuum variability in LRDs.","tokens_in":22186,"tokens_out":6544,"duration_ms":64259,"significance":"If the central prediction can be made robust, this paper offers a timely and physically motivated resolution to one of the key puzzles of LRD research: the absence of variability despite AGN-like spectra and compactness. Its strengths are the use of state-of-the-art radiation-MHD simulations to construct a variability model, the explicit comparison with published observational limits, and the provision of falsifiable predictions for the ongoing NEXUS and TWINKLE campaigns. The main weakness is that the super-Eddington optical PSD rests on a two-decade radial extrapolation from only two simulation anchors; the claimed non-detection fractions depend sensitively on this extrapolation. The qualitative conclusion that super-Eddington disks are less variable than DRW models is likely robust, but the quantitative consistency with current non-detections needs additional support.","major_comments":[{"comment":"The super-Eddington optical PSD is constructed by logarithmic extrapolation of beta and C0 from two inner-disk simulations (<10 r_g and 50-200 r_g) to 10^4 r_g. The two anchors already differ in slope (beta=2.4 vs 2.6) and normalization (C0=1e-6 vs 1e-7), so the extrapolated values (beta=3.0, C0=5e-9) have no direct support in the optical region. Section 4 acknowledges this but the alternative test (constant amplitude, tau proportional to r^3/2) only reduces variability further and does not bound the opposite direction, where a steeper or higher-normalized PSD at 10^4 r_g could raise the predicted detection fractions. Since the claimed <1% (KH24) and 12% (NEXUS) detection rates are controlled by these two numbers, the central quantitative claim is not yet robust. I recommend either obtaining an outer-disk anchor (e.g., from simulations at intermediate radii or from empirical constraints","section":"Section 2.2.2, Figure 1"},{"comment":"The DRW detection fractions depend on the assumed Eddington ratio L/L_Edd = 0.1 used to compute Mi from black hole masses. LRD Eddington ratios are uncertain and may be biased high if the objects are super-Eddington; the Burke et al. relation has a non-negligible dependence of SF_infinity on Mi (coefficient C=0.131). The paper should show how the predicted detection fractions (33-47% for T24/KH24) change when this assumption is varied over a plausible range (e.g., 0.01 to 1, or higher). The qualitative contrast with the super-Eddington model would likely survive, but the quantitative rejection of sub-Eddington models is only as strong as this assumption.","section":"Section 2.1, Eq. (1)"}],"minor_comments":[{"comment":"The title contains a typo: 'V ary' should be 'Vary'.","section":"Title"},{"comment":"The term 'structure function' is used for the PSD normalization C0, but C0 is not the structure function; please clarify the connection or use consistent terminology.","section":"Section 2.2.2"},{"comment":"The implementation of the low-frequency break is not fully described. Setting the light-curve length to the damping timescale in the Timmer & Koenig (1995) method effectively imposes a low-frequency cutoff, but this is not equivalent to a bending-power-law PSD with a break; please specify the exact procedure used.","section":"Section 2.2.2"},{"comment":"It would be helpful to explicitly state the assumed 3-sigma detection limit (Delta m = 0.21 mag) in the caption, and to differentiate it from the higher T24 limit shown as the thin gray line.","section":"Figure 2 caption"},{"comment":"The sentence 'both studies do not observe significant variability for any LRDs' is slightly ambiguous; T24 reported a mean Delta m that is not significant, but individual objects may show larger changes. Please clarify.","section":"Section 3.1.1"},{"comment":"For reproducibility, please provide a table of the input parameters used for the mock light curves (masses, redshifts, observing cadences, number of realizations) and consider releasing the analysis code or a data repository.","section":"Methods (global)"}],"recommendation":"major_revision","confidential_remarks":"The manuscript is well within the scope of this journal and addresses a high-profile puzzle with an interesting new model. The main issue is the load-bearing PSD extrapolation; I would like the authors to add sensitivity tests or find an empirical/simulational anchor before publication. The paper does not appear to have citation or novelty concerns, and the authors are appropriately transparent about their caveats."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"This paper is worth a look. It attacks the LRD variability puzzle by generating mock light curves from two models: the standard DRW prescription for sub-Eddington AGN, and a super-Eddington model built on radiation MHD simulations from Jiang et al. It makes concrete predictions for the NEXUS and TWINKLE surveys. The punchline is that DRW says we should have seen variability in many LRDs already, while the super-Eddington model says most should stay quiet. That comparison is genuinely new, at least in this form.\n\nWhat it does well: the methodology is transparent. The DRW implementation follows Burke et al. and uses EzTao; the super-Eddington PSD is anchored to actual simulations, and the authors explicitly list the assumptions. They also test one alternative scaling (constant amplitude, timescale scales as r^1.5), which pushes the predicted variability even lower. That is honest and useful. The qualitative conclusion is robust: if you take the models at face value, the observed non-detections are much more consistent with super-Eddington.\n\nThe load-bearing assumption is the extrapolation of the PSD from 10-200 r_g to 10^4 r_g. That is a two-decade leap in radius, and the stress-test note is right that the alternative scaling tested only makes variability smaller; it doesn't bound the case where the optical PSD is steeper or has higher normalization, or where irradiation plays a role. If the true PSD at 10^4 r_g is closer to nu^-2 with C0 ~ 1e-7, the predicted NEXUS detection fractions could go up substantially. The authors note this in Section 4, but they don't explore the opposite direction. Also, the super-Eddington model assumes UV variability is fully suppressed and that X-rays don't drive optical variability. These are plausible for the scenario they envision, but they are assumptions. The mass dependence is also speculative: they adopt a scaling from DRW without strong justification.\n\nNone of this kills the paper. The DRW comparison alone supports the conclusion that standard sub-Eddington models are inconsistent with the data. The super-Eddington model is a reasonable explanation, and the survey predictions are clearly stated. The caveats are disclosed, and the central argument is coherent.\n\nI'd send this to a serious referee. It deserves careful scrutiny, and the referee can push on the extrapolation and the assumptions. The paper is honest, clearly written, and will be cited by the JWST community. Conditional accept in my mind, not a reject.","headline":"A useful, transparent model comparison that makes the super-Eddington explanation for LRD non-variability plausible, though the optical PSD extrapolation is a real weakness.","tokens_in":22701,"tokens_out":2202,"would_cite":true,"duration_ms":21489,"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 argues that the absence of variability in little red dots is not a mystery to be explained away but a predicted signature of super-Eddington accretion: standard sub-Eddington AGN variability models, applied to the same observing c","keywords":["little red dots","super-Eddington accretion","AGN variability","damped random walk","radiation MHD simulations","JWST","high-redshift AGN","broad line variability"],"falsifier":"The NEXUS campaign results: the paper's super-Eddington model predicts that only ~12% of mock light curves reach Δm > 0.21 mag over its three-year baseline, while the sub-Eddington model predicts 97–100% do. If NEXUS detects strong continuum variability in a large majority of LRDs, the super-Eddington model fails; if nearly all stay quiet, the sub-Eddington model fails. A second, more immediate check is the paper's own alternative scaling (constant variability amplitude with timescales growing as r^1.5), which predicts magnitude changes two orders of magnitude smaller still and could be tested","tokens_in":21657,"feed_emoji":"🕳️","tokens_out":9131,"duration_ms":84807,"temperature":0.7,"pith_summary":"This paper asks why little red dots — the compact, red, V-shaped-spectrum objects JWST finds at high redshift, thought to be active galactic nuclei — barely change in brightness even when observed several times. The authors generated mock light curves with two models: the standard empirical 'damped random walk' description of variability calibrated on lower-redshift, slowly accreting (sub-Eddington) AGN, and a model built from radiation magnetohydrodynamic simulations of disks accreting a few times faster than the Eddington limit, the rate at which radiation pressure would balance gravity. With the sub-Eddington model, a third to a half of already-observed LRDs should have shown clear variability; none did. With the super-Eddington model, fewer than 1% of simulated light curves show detectable changes at current cadences, matching the observations. If the paper is right, the stillness of LRDs is expected for super-Eddington accretion rather than evidence against black-hole power, and two ongoing JWST campaigns, NEXUS and TWINKLE, can decide between the two models.","feed_headline":"Super-Eddington accretion explains why little red dots don't vary","feed_subtitle":"If little red dots are super-Eddington AGN, their stillness is exactly what the models predict.","key_machinery":"The load-bearing object is the power spectral density (PSD) of the disk's light curves. For the super-Eddington case, the authors fit power laws PSD ∝ C₀ ν⁻ᵝ to radiation MHD simulations of the inner disk — the soft X-ray region at ~10 r_g (β = 2.4) and the UV region at <200 r_g (β = 2.6) — and logarithmically extrapolate the fitted slope and amplitude to the optical photosphere at 10^4 r_g, obtaining β = 3.0 and C₀ = 5×10⁻⁹. That steep, low-amplitude PSD places essentially all optical variability at timescales of years or longer, far beyond the month-scale baselines of current and NEXUS observations. The contrasting machinery is the damped random walk (PSD ∝ ν⁻² above a damping timescale),","core_discovery":"The paper's central claim is that moderately super-Eddington accretion can easily account for the observed lack of variability in little red dots, while standard sub-Eddington AGN variability models cannot. For the sub-Eddington case the authors adopt the empirical damped-random-walk prescription from lower-redshift AGN studies and generate thousands of mock light curves sampled at the cadences of real observations; these predict detectable (>3σ) magnitude changes in 33–47% of the LRDs already studied, yet none of those objects show such changes. For the super-Eddington case they build a power-law power spectral density by extrapolating radiation MHD simulations from the inner disk (soft X-r","pith_inferences":["A consequence the paper leaves implicit: variability-selected searches for high-redshift AGN are biased toward sub-Eddington accretors, so a predominantly super-Eddington population would be systematically underrepresented in such surveys.","The inner-to-outer disk PSD extrapolation, if validated by NEXUS, could serve as a template for predicting variability in other X-ray-weak, gas-enshrouded AGN scenarios such as black-hole-star and quasi-star models.","A testable extension available right now: stacking all multi-epoch JWST photometry of LRDs to check whether the two-epoch magnitude-difference distribution is concentrated below ~0.1 mag with a thin large-variation tail, exactly as the super-Eddington model predicts.","If TWINKLE sees line variability in objects where NEXUS sees no continuum variability, that combination would constrain the irradiation geometry — soft X-rays reaching the broad line region while the disk photosphere stays shielded — a joint constraint no single observation currently provides."],"forward_implications":["If LRDs are moderately super-Eddington, their lack of variability is expected, and the AGN interpretation does not need exotic non-AGN alternatives to explain the quiet light curves.","The NEXUS campaign will discriminate: large magnitude changes (Δm > 1) in many LRDs would support sub-Eddington models, while near-universal quiescence, with only ~12% of sources above the 3σ threshold, supports super-Eddington accretion.","TWINKLE can detect broad emission line variability even when the continuum stays constant, provided soft X-ray emission from the inner disk reaches the broad line region — an indirect probe of X-rays that are not directly detected.","Rest-frame multi-year baselines, such as ten years of observer-frame monitoring or lensed LRDs with long time delays, should eventually reveal super-Eddington variability because the steep PSD predicts variability on few-year timescales.","The rare LRDs that do vary (for instance the two sources with correlated multi-band variability in the largest compiled sample) are consistent with the super-Eddington model's predicted small fraction of large excursions."],"supporting_citations":[{"why":"Supplies the structure-function scaling (SF∞ versus wavelength, magnitude, and black hole mass) used in Equation 1 for the sub-Eddington DRW predictions.","marker":"[Burke et al. 2023]"},{"why":"Supplies the mass-dependent damping timescale (Equation 2) that sets the characteristic timescale of DRW variability.","marker":"[Burke et al. 2021]"},{"why":"Radiation MHD simulations of the UV-emitting region (50–200 r_g) of super-Eddington disks; source of the UV PSD used in the optical extrapolation.","marker":"[Jiang et al. 2025]"},{"why":"Simulations of the soft X-ray emitting inner disk (~10 r_g); provides the shallower PSD anchor for the radial extrapolation.","marker":"[Jiang et al. 2019]"},{"why":"Supplies the 30-LRD two-epoch HST+JWST variability dataset whose cadence and null result the mock light curves must reproduce.","marker":"[T24]"},{"why":"Supplies the multi-epoch JWST observations and observing times for example LRDs used to define the mock sampling.","marker":"[KH24]"},{"why":"Compiles 314 LRDs with multi-epoch JWST data, establishing the population-wide lack of variability that both models are judged against.","marker":"[Zhang et al. 2024]"},{"why":"Provides the standard method for generating mock light curves from a specified power spectral density, used for both models.","marker":"[Timmer & Koenig 1995]"},{"why":"Shows that X-ray irradiation is required to drive UV/optical variability in disk light curves, justifying the super-Eddington assumption of purely local intrinsic variability.","marker":"[Secunda et al. 2025]"},{"why":"Defines the NEXUS campaign whose anticipated cadence and baseline anchor the paper's future variability predictions.","marker":"[Shen et al. 2024]"}],"fun_headline_variants":["Super-Eddington disks explain why little red dots don't flicker","Little red dots' calm hides super-Eddington accretion","Why JWST's little red dots stay still: super-Eddington disks","Super-Eddington model matches little red dots' lack of variability","No wiggle for little red dots: super-Eddington accretion"],"cache_read_input_tokens":2688,"weakest_assumption_plain":"The entire super-Eddington prediction rests on one long extrapolation: the variability scaling measured between 10 and 200 r_g in the simulations is assumed to hold unchanged out to the optical emitting region at 10^4 r_g, a step the paper itself flags in Section 4 as the model's main uncertainty.","fun_headline_variants_meta":{"raw":{"variants":["Super-Eddington disks explain why little red dots don't flicker","Little red dots' calm hides super-Eddington accretion","Why JWST's little red dots stay still: super-Eddington disks","Super-Eddington model matches little red dots' lack of variability","No wiggle for little red dots: super-Eddington accretion"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000162,"raw_usage":{"total_tokens":1121,"prompt_tokens":834,"completion_tokens":287,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":578,"completion_tokens_details":{"reasoning_tokens":205}},"tokens_in":578,"tokens_out":287,"duration_ms":3115,"temperature":1.0,"reasoning_tokens":205,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-05T10:50:04.900644+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"The NEXUS campaign results: the paper's super-Eddington model predicts that only ~12% of mock light curves reach Δm > 0.21 mag over its three-year baseline, while the sub-Eddington model predicts 97–100% do. If NEXUS detects strong continuum variability in a large majority of LRDs, the super-Eddington model fails; if nearly all stay quiet, the sub-Eddington model fails. A second, more immediate check is the paper's own alternative scaling (constant variability amplitude with timescales growing as r^1.5), which predicts magnitude changes two orders of magnitude smaller still and could be tested","supporting_citations":[{"cited_title":"1995, , 300, 707","cited_arxiv_id":null,"evidence_quote":"Provides the standard method for generating mock light curves from a specified power spectral density, used for both models."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Shows that X-ray irradiation is required to drive UV/optical variability in disk light curves, justifying the super-Eddington assumption of purely local intrinsic variability."}],"review_version":1}