{"id":"70ee8ada-8576-44fb-965d-7ab0081e678d","arxiv_id":"2601.15962","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":5.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":7,"one_line_summary":"Faint AGN hosts at z≈4–6 look 1–2 dex less massive after image decomposition, pushing their black-hole-to-stellar-mass ratios to ~0.01–1.48, above the local relation.","lead":"JWST image-fitting of 17 faint AGN at z≈4–6 finds that, after subtracting the bright black-hole point source, the galaxies hosting them are 1–2 dex less massive than photometry-only fits suggest. If the black-hole masses are right, these systems sit far above the local black-hole–galaxy mass relation—or their hosts are too compact to resolve.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Elevated M_BH/M* hinges on uncalibrated single-epoch virial masses; the 2-dex electron-scattering alternative would erase the offset.","rationale":"The paper is internally consistent and unusually transparent; the mock tests in Appendix A support host-flux recovery for AGN/host ratios below ~1000, and the size–mass agreement supports the host interpretation. However, the headline M_BH/M* ratio inherits the poorly calibrated single-epoch virial scale. The paper's own §5.4 admits a 2-dex downward correction is possible, and that is larger than the entire claimed offset above the local relation. The reader already assigned CONDITIONAL on this basis; my stress test therefore does not move the verdict. The 15-vs-9 host-detection inconsistency in Table 3 is worth fixing but is secondary because the BH-mass systematic dominates the ratio. A direct sensitivity calculation—recompute with 2-dex-lower M_BH—would settle whether the conclusion survives.","tokens_in":25848,"tokens_out":6272,"duration_ms":60545,"concrete_test":"Recompute Table 3 and Fig. 6 using M_BH values derived under the electron-scattering prescription cited in §5.4 (or, as a bounding case, with all M_BH reduced by 2 dex) while keeping the reported host stellar masses. If the secure 9-source subset no longer lies above the local M_BH–M* relation, the central claim is not robust to the dominant systematic. A complementary direct check: obtain one dynamical BH-mass measurement for a secure host in this sample (e.g., via NIRSpec IFU or ALMA gas kinematics) and compare it with Eq. 2.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central result—M_BH/M* = 0.01–1.48, well above local relations—uses black-hole masses from Eq. 2 (J. Zhang et al. 2025), a locally calibrated single-epoch virial estimator that assumes the broad Hα FWHM is entirely Doppler. §5.4 explicitly concedes that no direct calibration exists beyond z≳2 and that, under the electron-scattering model of Rusakov et al. 2025, BH masses could be ~2 dex lower. This is a factor-of-100 systematic on the numerator of the headline ratio: if it applies, the claimed excess largely disappears. The paper cites one dynamical LRD mass (Juodžbalis et al. 2025b) as support, but that measurement is for a different object, not for this sample, and Eq. 2 remains unvalidated at z=4–6. The imaging-side inconsistency—Table 3 labels 15 sources as host detections while only 9 are secure (ΔBIC<−500)—is real but secondary; it does not remove the BH-mass issue. Since the authors themselves flag this limitation, the verdict should remain conditional, not accept.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper performs AGN-host image decomposition for 17 broad-Hα-selected AGN at z≈3.7–6.5 in the JADES GOODS-N field, using galfit+MCMC on up to seven NIRCam wide-band images. The central claims are: (1) extended host-galaxy emission is robustly detected in 9 of 17 sources; (2) when image decomposition is used before SED fitting, host stellar masses can be 1–2 dex lower than those from photometry-only SED fitting; (3) the resulting M_BH/M* ratios, 0.01–1.48, place these objects well above the local M_BH–M* relation; and (4) the hosts may be genuinely undermassive or too compact to be resolved. The paper also presents mock tests in Appendix A assessing host flux recovery and cross-checks against an independent measurement of GN53757.","tokens_in":26215,"tokens_out":3984,"duration_ms":42122,"significance":"If correct, the paper provides one of the first systematic imaging-decomposition constraints on host stellar masses for z≈4–6 broad-line AGN and highlights a real methodological issue: photometry-only SED fits can overestimate host masses by 1–2 dex. The mock experiments, cross-validation, and public data products are useful. However, the headline M_BH/M* offset is directly contingent on single-epoch virial BH masses adopted from the companion paper (J. Zhang et al. 2025), whose Doppler-broadening assumption is explicitly acknowledged as uncalibrated beyond z≈2 and as potentially overestimating M_BH by up to 2 dex under electron scattering. That factor of 100 on the numerator of the central ratio is load-bearing. In addition, the paper conflates the 9 secure host detections (ΔBIC<−500) with 15 sources listed in Table 3 as 'host detection,' mixing marginal and strong detections in the mass-ratio analysis. The work is therefore a valuable contribution to a rapidly moving field, but the central claim needs a quantitatively framed robustness analysis before it can be accepted at face value.","major_comments":[{"comment":"The central M_BH/M* ratios are computed from BH masses in Table 3 that rely entirely on single-epoch virial masses from J. Zhang et al. (2025) using Eq. (2). As the text itself states, this estimator assumes the broad Hα FWHM is purely Doppler and has no direct calibration at z≳2. Under the electron-scattering scenario of Rusakov et al. (2025), M_BH could be up to ~2 dex lower, which would erase the claimed offset from the local M_BH–M* relation for most of the sample. This is not a peripheral caveat; it is the numerator of every ratio in Figure 6. The paper should present a quantitative alternative: e.g., recompute the offset and the fraction of objects above local relations with M_BH shifted by −2 dex, or provide an additional BH-mass estimate with an explicit uncertainty budget that includes this systematic. Without this, the headline 'undermassive hosts' claim is conditional on an un","section":"§5.4, Eq. (2)"},{"comment":"Table 3 is titled 'AGN galaxies with host detection' and includes 15 sources, but the paper's own criterion in Table 1 flags only 9 as secure (ΔBIC<−500; flag 'Y'). The other six have −500<ΔBIC<0 (flag 'X') and are explicitly described as lacking clear visual confirmation in residual images. The abstract and conclusions correctly state 9 detections, yet the M_BH/M* analysis appears to use all 15 entries. This mixes secure and marginal detections in the central result. Please present the 9-source secure subset separately, and treat the six marginal cases as tentative/upper limits with a flag in Table 3 and Figure 6.","section":"Table 3, §4.1, §5.4"},{"comment":"The mock tests validating host-flux recovery are built on a single compact template galaxy (GN1034159) and assume that a wavelength-independent Sérsic profile (fixed n and R_e across F090W–F444W) describes all targets. The paper notes that GALFITM fits are unstable in F356W/F444W, but the representativeness of a one-template test is not demonstrated. More importantly, the mock conclusions are stated as 'reliable recovery' for F090W–F277W under AGN/host<1000 and SNR≳3, while F356W/F444W are unreliable at low SNR; the SED fits that yield stellar masses include these LW bands. The paper should either propagate the LW flux uncertainties into the final stellar-mass errors or demonstrate explicitly that the derived M* values are insensitive to the unreliable LW bands.","section":"Appendix A, §3"}],"minor_comments":[{"comment":"The abstract says 'z∼4–6' while Section 6 states 'z=3.7–6.5'. Please harmonize the redshift range.","section":"Abstract vs. §6"},{"comment":"The stated M_BH/M* range is 0.01–1.48, but Table 3 gives log(M_BH/M*)=−2.04 for GN1090253 (ratio≈0.009) and −1.93 for GN1082263 (ratio≈0.012). The text's lower bound should be consistent with the table.","section":"§5.4"},{"comment":"The 'Filter' column is the band used for the representative Sérsic profile, but the table caption does not explicitly say so; adding '(representative SW band)' would improve clarity.","section":"Table 1"},{"comment":"The discussion of nebular emission as an alternative interpretation is balanced, but the sentence 'the non-detection of He ii λ1640 and Mg ii λλ2796,2803...' would benefit from a citation or a note that these are from the CONGRESS/FRESCO spectra presented in the companion paper.","section":"§5.1"}],"recommendation":"major_revision","confidential_remarks":"The paper is timely and the imaging-decomposition method is generally well executed, with useful mock tests and cross-checks. My main concern is the asymmetric treatment of the BH-mass systematic: the authors acknowledge that the electron-scattering scenario could reduce the central ratio by 2 dex but do not incorporate that into the reported uncertainties or conclusions. The Table 3 labeling of 15 sources as host detections when only 9 are secure is another fixable but important inconsistency. These are fixable within the scope of a revision, so I recommend major revision rather than rejection."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Bottom line: this is a useful and unusually candid paper. The new data are host-galaxy fluxes, stellar masses, and sizes for 17 CONGRESS/FRESCO broad-H-alpha AGN at z~4–6, obtained with GALFIT+MCMC decomposition of JADES NIRCam images. The main claim that photometry-only SED fitting overestimates host stellar masses by 1–2 dex when the AGN component is not recognized is well supported by their mocks and by the internal comparison to the companion SED fits. The elevated M_BH/M* ratios are real under the stated assumptions, but they depend on BH masses from J. Zhang et al. (2025) using a locally calibrated single-epoch virial estimator with no direct calibration beyond z~2. The paper itself concedes in §5.4 that the electron-scattering alternative could lower BH masses by ~2 dex, which would largely erase the offset. That is a load-bearing caveat, and they are honest about it.\n\nWhat the paper does well: the mock experiments in Appendix A are thoughtful and actually constrain where the method breaks down (SW bands good, F356W/F444W unreliable at low SNR). The validation against an independent Forcepho fit for GN53757 is a genuinely useful cross-check. The size–mass relation agreement adds credibility to the interpretation that the extended component is stellar. The writing is clear about the nebular-emission alternative and about foreground interloper contamination in two sources.\n\nSoft spots, in proportion: the biggest issue is not the method but the framing. Table 3 presents stellar masses for 15 sources as \"host detection,\" while the text says only 9 are secure (ΔBIC<−500) and 6 are tentative (ΔBIC between −500 and 0). The table does not flag which rows are tentative, so a reader can easily overcount the secure sample. That should be fixed by adding flags or splitting the table. The reported uncertainties on radius and Sérsic index are statistical only; the mocks show wavelength-dependent systematic biases, so the error bars understate the true uncertainty. None of these are fatal, and the central host-mass conclusion is robust to them. The M_BH/M* conclusion, however, is conditional on the BH masses, and the paper would be stronger if that conditionality were in the abstract.\n\nThis paper deserves a serious referee. It is not a methodological breakthrough — the approach closely follows Chen et al. (2024) and Harikane et al. (2023) — but the sample is new, the comparison to photometry-only SED fitting is systematic, and the caveats are handled with unusual honesty. I would send it to review with a request to clarify the detection tiers in Table 3 and to frame the M_BH/M* result as explicitly contingent on the virial BH-mass assumption. I would cite this for the host masses of faint high-z AGN.","headline":"A solid, honest imaging-decomposition study of 17 faint z~4–6 AGN; the host-mass result is well supported, but the headline M_BH/M* offset inherits uncalibrated single-epoch BH masses that the authors themselves flag could erase the offset.","tokens_in":26798,"tokens_out":1422,"would_cite":true,"duration_ms":15466,"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":"By separating AGN light from host starlight in JWST/NIRCam images of 17 faint AGN at z≈4–6, this paper finds hosts 10–100 times lighter than photometry-only fits imply, driving black-hole-to-stellar-mass ratios up to 1.48 and far above the","keywords":["active galactic nuclei","supermassive black holes","host galaxy stellar mass","black hole–stellar mass relation","high-redshift galaxies","image decomposition","broad Hα AGN","JWST/NIRCam"],"falsifier":"A reverberation-mapping or direct dynamical measurement for one of the nine hosts with detected extended light—using time-resolved or spatially resolved spectroscopy of the broad Hα line—that yields a black-hole mass roughly 2 dex below the single-epoch virial estimate would falsify the elevated MBH/M* ratios; conversely, resolving a compact stellar component that accounts for the missing 1–2 dex in stellar mass would falsify the under-massive-host interpretation.","tokens_in":25735,"feed_emoji":"🌌","tokens_out":8195,"duration_ms":82920,"temperature":0.7,"pith_summary":"This paper asks whether the tight local link between supermassive black hole mass and host galaxy stellar mass already exists at z≈4–6, or whether black holes assemble first. The authors decompose JWST/NIRCam images of 17 faint, broad-Hα-selected AGN into a point-source component (the AGN) and a Sérsic-profile component (the host), then fit the separated host fluxes to measure stellar masses. They robustly detect extended host light in 9 of 17 sources; for these, stellar masses come out 1–2 dex (10–100×) lower than masses from photometry-only SED fits that fail to identify the AGN. The resulting MBH/M* ratios span 0.01–1.48, well above the local relation, implying either genuinely under-massive hosts or hosts too compact to resolve. The result matters because it suggests black-hole growth can outpace host-galaxy assembly in the early universe, and it shows why spatial decomposition is necessary before quoting stellar masses of JWST-discovered AGN.","feed_headline":"JWST finds early black holes are ~1000x overmassive for their hosts","feed_subtitle":"Separating AGN from host-galaxy light in JWST images shows hosts 10-100x lighter than photometry-only fits.","key_machinery":"The central machinery is a two-component image decomposition: each AGN is modeled as a point spread function, and the host as a single Sérsic profile convolved with the same PSF. Structural parameters (effective radius, Sérsic index) are taken from the best short-wavelength band where the host is least outshone, then held fixed while Markov-chain Monte Carlo fitting scales the PSF and Sérsic fluxes in all seven NIRCam bands; AGN-subtracted images are used for aperture photometry, and the resulting host SED is fit to derive stellar mass. Host-detection significance is quantified by the Bayesian Information Criterion difference between a PS-only and a PS+Sérsic model, with ΔBIC<−500 treated as","core_discovery":"The core claim is that for 17 broad-line AGN at z≈3.9–5.5, spatially decomposing AGN and host light with a PSF-plus-Sérsic model changes inferred stellar masses dramatically: when the photometry-only SED fit had not recognized the AGN, the host masses are ~0.9–1.8 dex lower. With those corrected masses, the black-hole-to-stellar-mass ratios span 0.01–1.48, placing the sample above the local MBH–M* relation by roughly a factor of 100–1000. The paper argues the extended light is the host galaxy because its centroid lies within ~0.12 kpc of the AGN, its size–mass relation matches normal galaxies, and its detection rate (~53%) is consistent with other JWST findings; it leaves open that some host","pith_inferences":["If electron scattering substantially broadens the Hα line, the black-hole masses would fall by up to ~2 dex and the apparent ratio excess would largely vanish; this is an inference beyond the paper, which only flags the possibility and notes the lack of direct calibration beyond z≈2.","The paper's two readings—under-massive hosts versus unresolved compact hosts—make opposite predictions for high-resolution follow-up: deeper or longer-wavelength imaging should reveal either a faint extended stellar disk or a compact stellar concentration hidden in the PSF.","The 53% host-detection fraction may reflect a selection effect: hosts that are fainter relative to the AGN are harder to detect, so the most extreme ratios could be preferentially missed or preferentially spurious.","If these ratios are real, they align with the most extreme high-redshift systems reported elsewhere, suggesting that a black-hole-first growth phase may be common in the early universe and that heavy-seed formation channels deserve closer scrutiny—though the virial-mass caveat makes this inference conditional."],"forward_implications":["If the decomposition is correct, photometry-only SED fits that miss the AGN overestimate host stellar masses by 1–2 dex, so the faint-AGN host population at z≈4–6 is substantially lighter than previously inferred.","These systems sit a factor of ~100–1000 above the local black-hole–stellar-mass relation, with MBH/M* ratios from 0.01 to 1.48.","The detected hosts follow the normal galaxy size–mass relation, so the extreme ratios are not caused by unusually large or diffuse hosts.","Only 9 of 17 sources show securely detected extended emission; for the remaining sources the light is consistent with a pure point source, so their mass ratios are either lower limits or their hosts are too compact to be seen.","If some host starlight is unresolved in the central point source, the true stellar masses would be higher, but the systems would still point to inside-out growth: black holes assembling before or faster than their hosts."],"fun_headline_variants":["JWST: early black holes 100-1000x overmassive vs host galaxies","Splitting AGN from host light reveals overmassive black holes at z~5","Early AGN hosts 10-100x lighter than inferred without decomposition","Black holes at z~4-6 outweigh hosts by up to 1000x in JWST data"],"cache_read_input_tokens":2304,"weakest_assumption_plain":"The load-bearing premise is that the measured width of the broad Hα line is entirely Doppler broadening from virial motion, so the single-epoch black-hole mass estimate is correct; the paper itself notes that no direct calibration exists beyond z≈2 and that electron scattering could lower the masses by up to ~2 dex, which would erase most of the apparent ratio excess.","fun_headline_variants_meta":{"raw":{"variants":["JWST: early black holes 100-1000x overmassive vs host galaxies","Splitting AGN from host light reveals overmassive black holes at z~5","Early AGN hosts 10-100x lighter than inferred without decomposition","Black holes at z~4-6 outweigh hosts by up to 1000x in JWST data"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000739,"raw_usage":{"total_tokens":3229,"prompt_tokens":928,"completion_tokens":2301,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":672,"completion_tokens_details":{"reasoning_tokens":2207}},"tokens_in":672,"tokens_out":2301,"duration_ms":20102,"temperature":1.0,"reasoning_tokens":2207,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-03T08:42:12.080657+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"A reverberation-mapping or direct dynamical measurement for one of the nine hosts with detected extended light—using time-resolved or spatially resolved spectroscopy of the broad Hα line—that yields a black-hole mass roughly 2 dex below the single-epoch virial estimate would falsify the elevated MBH/M* ratios; conversely, resolving a compact stellar component that accounts for the missing 1–2 dex in stellar mass would falsify the under-massive-host interpretation.","supporting_citations":[],"review_version":1}