{"id":"f0192e4a-542b-4bcd-be52-cb9ae74e43e4","arxiv_id":"2411.14535","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":4.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":3,"one_line_summary":"Ten dwarf galaxies in the local universe host radio-excess active galactic nuclei selected from VLA Sky Survey data, five of which are new AGN identifications.","lead":"Using a radio-infrared excess cutoff derived from about 7,000 galaxies, the authors identify 10 dwarf galaxies whose radio emission is too strong to come from star formation alone, likely powered by active black holes. Five of these AGN candidates in dwarf galaxies are new, and a public catalog of 6,904 radio-excess AGN hosts is provided.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The q<1.94 cutoff is calibrated on W4 S/N≥5 point-like sources but applied to W4-faint, potentially extended dwarfs; underestimated W4 fluxes can create false radio-excess AGNs.","rationale":"The central claim depends on the q threshold doing the same work for dwarf galaxies as for the calibration sample. The calibration is explicitly restricted to W4 S/N≥5 point-like sources; the application to dwarfs removes those restrictions. Since profile-fit W4 photometry is known to underestimate extended sources, and nearby dwarfs can be resolved at 12″, the transfer has a concrete bias direction: LTIR too low, q too low, enhanced false-positive rate. The authors' appeal to Delvecchio et al. addresses the mean q, not the scatter or the photometric bias. The background-contamination estimate is also a concern, but even if several of the ten were background, the majority-association claim could survive; the q-selection failure would overturn the classification of the five new AGN candidates and is therefore more load-bearing. The paper is otherwise careful: catalogs, spectral indices, SNR checks, and independent multiwavelength evidence for some objects are real strengths. A conditional acceptance is appropriate pending the photometric and statistical check.","tokens_in":33838,"tokens_out":15481,"duration_ms":163909,"concrete_test":"Retrieve AllWISE w4chi2 and use unWISE or forced-aperture photometry to recompute LTIR for the 10 candidates and for the radio-detected SF-consistent dwarfs. Count how many candidates with w4chi2>3 or with W4 upper limits move above q=1.94 when total, extension-safe fluxes are used; then measure the scatter of q for the SF dwarf sample and compute the expected number of SF dwarfs below q=1.94. If any candidate no longer satisfies q<1.94, or if the expected SF leakage exceeds about one object, the ten-candidate claim needs revision.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Section 3.1 defines the calibration sample with W4 S/N≥5 and discards 219 extended WISE sources (w4chi2>3) because profile-fit photometry underestimates fluxes of resolved sources. Section 5.1 then applies the resulting q<1.94 cutoff to dwarf galaxies without either a S/N cut or an extended-source correction. Many dwarfs at z≤0.15 have angular sizes comparable to the 12″ W4 PSF, so their W4 profile-fit fluxes—and hence LTIR in Equation (1)—can be systematically low. For 7 of the 10 candidates, log LTIR is an upper limit, so the reported q is an upper limit; if total or forced-aperture photometry gives a larger LTIR, q may exceed 1.94. The paper acknowledges uncertainty in applying the high-mass cutoff to dwarfs and cites Delvecchio et al. for the mean q rising at low stellar mass, but it does not quantify the dwarf q distribution or the false-positive fraction from star-forming dwarfs. The additional photometric bias from applying a point-source-calibrated selection to potentially resolved dwarfs makes the transfer even less secure. The five newly identified dwarf-AGN claims rest entirely on this unquantified selection-function premise.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper searches for radio-excess AGN candidates in dwarf galaxies by defining an infrared-radio correlation (IRRC) parameter q from VLASS 3 GHz and WISE W4 22 micron measurements. A Gaussian fit to the q distribution of ~6,800 galaxies in the NASA-Sloan Atlas with good W4 detections gives q_peak = 2.62, sigma = 0.34, and a 2-sigma cutoff q < 1.94. This cutoff is applied to 63,656 NSA dwarfs (M* <= 3e9 Msun, z <= 0.15) cross-matched to VLASS, yielding 123 radio matches, then 50 bona fide dwarfs after visual and spectroscopic interloper removal (HII regions, quasars, an unreliable NSA mass, PTF11qcj, and the background quasar J1136+1252). Ten objects with q < 1.94 are presented as the final radio-excess AGN sample, with five claimed as newly identified AGNs in dwarfs and eight variable radio sources relative to FIRST or VLASS epoch 1. The authors provide public catalogs of radio-excess AGNs and SF-consistent galaxies, and they argue on statistical grounds and via emission-line diagnostics that most of the ten sources are physically associated with their dwarf hosts.","tokens_in":1986,"tokens_out":1943,"duration_ms":35378,"significance":"If the ten candidates are genuine dwarf-galaxy AGNs, the paper provides a valuable, homogeneous VLASS-based sample of IMBH candidates, including several objects not previously recognized as AGNs and eight variables that merit follow-up. The publicly released catalogs (radio-excess AGNs and SF-consistent galaxies) are a useful community resource, and the careful interloper removal and quantitative checks against SNR luminosity relations are strengths. The novelty is incremental relative to Reines et al. (2020) and other radio searches, but the extension to the larger NSA volume with VLASS yields new candidates and demonstrates a reproducible selection recipe. The central risk is the transfer of the q < 1.94 threshold from a W4-bright, point-like calibration sample to W4-faint, potentially extended dwarfs, which is the main correctness concern for the headline claim; the paper itself acknowledges this uncertainty only qualitatively.","major_comments":[{"comment":"The q < 1.94 cutoff is calibrated on galaxies with W4 S/N >= 5 and on point-like WISE sources (the 219 extended sources with w4chi2 > 3 are removed), but it is then applied to dwarf galaxies without either a W4 S/N cut or an extended-source correction. Seven of the ten candidates have upper-limit L_TIR (Table 1), so their q values are formally upper limits; if W4 profile-fit fluxes are underestimated for resolved dwarf hosts, the true q may exceed 1.94 and the radio-excess classification could fail. The authors discuss this selection-function issue qualitatively (including the Delvecchio et al. 2021 result that <q> rises at low stellar mass and their own statement that the dwarf cutoff is likely higher), but they do not quantify the dwarf q distribution, the expected false-positive rate from star-forming dwarfs, or the size of the W4 extended-source bias. Since the five newly identified dwarf-AGN claims rest entirely on this transfer, a quantitative treatment (or a conservative re-analysis with forced aperture photometry or W4 upper limits treated as full limits) is needed.","section":"Section 3.1 and Section 5.1"},{"comment":"The claim that candidate IDs 6 and 7 are 'extended' (Psi_maj of 3.2 and 2.6 arcsec, respectively) is presented as possible jet evidence, but extended structure is also a natural consequence of the star-forming disk emission that the q criterion is meant to exclude. Given that most of the ten sources have upper-limit L_TIR and hence uncertain q, the two extended sources are the least secure AGN identifications; a VLASS quick-look image cutout with a resolved SF disk or a stacking test would clarify whether the extended radio emission is compact-core plus jets or distributed disk emission.","section":"Section 5.3 and Table 1"},{"comment":"The background contamination estimate of 4.4 +/- 2 coincidental matches is derived from the full 123 dwarf-VLASS matches and then quoted as implying 'less than 1' background source among the 10 final candidates. Because the final sample is filtered by q < 1.94 and by WISE associations, the background fraction among q-selected objects is not necessarily the same as the geometric expectation for all 123 matches; a background AGN that is radio loud and infrared faint will preferentially pass the q cut. The statistical argument would be more convincing if it were recomputed for the q < 1.94 subpopulation or with a Monte Carlo that includes realistic radio flux distributions of background AGNs.","section":"Section 5.2"},{"comment":"The comparison to Reines et al. (2020) correctly explains why only 2 of the 13 Reines et al. objects survive into the new sample, but the discussion treats the five VLASS non-detections of FIRST sources above 4 mJy as 'likely variable sources' without ruling out catalog-level issues such as VLASS quality flags, sidelobe confusion, or source blending at 2.5 arcsec resolution. A simple check of the VLASS quick-look images at the Reines et al. positions would strengthen the variability claim; as written, the variability interpretation is plausible but not fully demonstrated.","section":"Section 5.6"}],"minor_comments":[{"comment":"The phrase 'radio-excess AGNs' is used interchangeably for candidates and confirmed AGNs; the abstract should state explicitly that the ten are candidates requiring follow-up.","section":"Abstract and Section 1"},{"comment":"The processing description says the CIRADA quality flag removes sources with peak brightness less than 5 times the local RMS, but the paper does not state how the integrated flux (S_3GHz) in Table 1 is measured (CIRADA catalog values vs. new aperture photometry); a sentence on the provenance of the tabulated flux densities would remove ambiguity.","section":"Section 2.1"},{"comment":"The spectral-index conversion is applied to sources without FIRST detections assuming alpha = -0.7, but the text does not report the uncertainty introduced by this assumption into q; a brief error-propagation statement would help the reader gauge the 2-sigma threshold's robustness.","section":"Section 3.3, Equation 2"},{"comment":"The units of L_TIR are listed as W, while L_3GHz and L_1.4GHz are in W Hz^-1; this is correct but should be stated in the table note for clarity.","section":"Table 1"},{"comment":"The sentence 'we find considerable uncertainty in the average value of q for galaxies with masses M* <= 10^9.5 Msun' is understated; the small-number caveat applies to the entire dwarf mass regime and should appear earlier in the selection discussion.","section":"Section 5.1"},{"comment":"Several objects (IDs 3, 7, 9) have W3 upper limits and are plotted as triangles; the caption should note which points are upper limits in each band so the reader does not infer constraining WISE colors for those objects.","section":"Figure 19"},{"comment":"The variable-source count of 8 is assembled from heterogeneous criteria (objects detected in VLASS but not FIRST, objects detected in FIRST but not VLASS, and one object missing only VLASS epoch 1); a table or list of the 8 objects with their detection history (FIRST, VLASS e1, VLASS e2) would make the claim more verifiable.","section":"Section 5.7"}],"recommendation":"major_revision","confidential_remarks":"The paper is a solid observational search with careful interloper removal and useful public catalogs. My recommendation is driven by the unquantified transfer of the q cutoff from a W4-bright, point-like calibration sample to W4-faint, potentially extended dwarfs, which is the load-bearing step for the five new dwarf-AGN claims. I do not see this as a fatal flaw: the authors clearly recognize the issue and could address it with a focused analysis (e.g., forced-aperture or total W4 photometry for the ten candidates, a Monte Carlo false-positive estimate for star-forming dwarfs, or a recomputed background expectation for the q-selected subsample). The extended-source concern for IDs 6 and 7 is also addressable with existing VLASS images. No concerns about citation practice or scope beyond the normal fit of a survey-oriented AGN search paper to ApJ."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Two things to know about arXiv:2411.14535. It is a solid catalog paper: ten dwarf-galaxy radio-excess AGN candidates from VLASS, five of them new, plus a community catalog of 6,904 radio-excess AGNs. The main soft spot is that the q<1.94 cutoff is calibrated on W4-bright, point-like galaxies and then applied to dwarfs with W4 upper limits and no correction for resolved WISE sources; that transfer is not quantified.\n\nWhat's genuinely useful here: the VLASS epoch 2 data and NSA v1_0_1 extend the dwarf AGN search beyond Reines et al. (2020), and the small cross-match radius plus careful interloper removal (HII regions, quasars, PTF11qcj, the background quasar J1136+1252) show good judgment. The checks against SNR luminosity relations are a real plus: all ten sources sit well above the expected SNR population, so those are unlikely to mimic AGNs. The paper also does a thorough comparison to Reines et al., explaining why 11 of the 13 previously known sources are absent here — mostly due to variability, offset, or the VLASS catalog limits. That kind of documentation is valuable.\n\nThe fragile step is the selection function. The calibration sample in Section 3.1 requires W4 S/N≥5 and removes 219 extended WISE sources because the profile-fit photometry underestimates resolved fluxes. Section 5.1 then applies the resulting cutoff to dwarfs without a W4 S/N cut and without an extended-source correction. Seven of the ten candidates have LTIR as an upper limit. In the standard interpretation an upper limit means the true q is lower, so the object is even more radio-excess; but if the W4 flux is biased low because the dwarf is resolved, the true q could be higher than reported. The paper acknowledges the mass-dependent uncertainty and cites Delvecchio et al. showing q rises at low mass, so the cutoff is likely conservative in that respect — but it does not quantify the dwarf IRRC scatter or the false-positive rate from star-forming dwarfs. That is the main weakness, and it is addressable.\n\nTwo smaller issues: the background-contamination estimate (4±2 coincidental matches) is computed for all 123 radio sources, not specifically for the ten radio-excess candidates; and the paper does not tabulate q and its uncertainty for each candidate. Both would be easy to add.\n\nWho is this for? Anyone working on dwarf-galaxy AGNs or IMBH occupation fractions. The catalog itself is worth having, and the ten candidates are pointers for follow-up. I would send this to a referee. The referee should push for a dwarf-calibrated selection function or a quantitative estimate of the photometric bias, tabulated q values, and a background estimate restricted to the radio-excess set. With those, I'd be happy to see it published. If the authors can't tighten the selection-function transfer, the paper would still work as a catalog paper, but the 'five new dwarf AGNs' claim should be softened.","headline":"A careful, useful VLASS-based dwarf-galaxy AGN catalog whose main weakness is an unquantified transfer of the IR-radio cutoff to W4-faint, possibly extended dwarfs.","tokens_in":34678,"tokens_out":7512,"would_cite":true,"duration_ms":63838,"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":"Ten nearby dwarf galaxies host radio-excess AGNs picked out by a 2-sigma infrared-radio cutoff, and most are physically associated with their hosts.","keywords":["dwarf galaxies","radio-excess AGN","infrared-radio correlation","intermediate-mass black holes","VLASS","WISE","radio variability","active galactic nuclei"],"falsifier":"Measure the full $q$ distribution for a volume-limited sample of dwarf galaxies with no AGN signatures, using deep radio and infrared data; if more than about two percent of such dwarfs have $q<1.94$, the ten radio-excess AGN candidates could be ordinary low-$q$ star-forming galaxies rather than accreting black holes.","tokens_in":33609,"feed_emoji":"🔭","tokens_out":8039,"duration_ms":70029,"temperature":0.7,"pith_summary":"The paper tries to establish that a simple radio-versus-infrared ratio can pick out accreting massive black holes in dwarf galaxies, where optical and X-ray surveys miss many sources. Using roughly 7,000 galaxies with radio detections from VLASS and infrared detections from WISE, the authors calibrate the infrared-radio correlation parameter $q$ and set a $2\\sigma$ threshold of $q < 1.94$ to separate star-formation radio emission from radio-excess AGNs. Applying this threshold to dwarf galaxies with $M_\\star \\le 3\\times 10^9\\,M_\\odot$ and $z \\le 0.15$, after removing interlopers and checking supernova-remnant explanations, they find ten candidates, five of which are identified as AGNs for the first time. If the physical association with the hosts holds, these are new intermediate-mass black hole candidates in a mass range that is hard to populate.","feed_headline":"Ten dwarf galaxies host radio-bright black hole candidates","feed_subtitle":"Five of the ten were never flagged as AGNs before; the rest are new intermediate-mass black hole candidates.","key_machinery":"The load-bearing object is the infrared-radio correlation parameter $q$, defined by $q = \\log(L_{\\rm IR}/(3.75\\times 10^{12}\\,L_{1.4\\,{\\rm GHz}}))$, which captures the tight relation between radio and far-infrared emission in star-forming galaxies. The authors measure $q$ for a sample of about 7,000 galaxies drawn from the NASA-Sloan Atlas with VLASS 3 GHz detections, WISE W4 detections with $S/N \\ge 5$, and point-like WISE profiles, excluding mid-IR-selected AGNs. The distribution peaks at $q_{\\rm peak}=2.62$ with scatter $\\sigma = 0.34$, giving the $2\\sigma$ threshold $q<1.94$. This threshold is what separates star-formation-powered radio emission from radio-excess AGNs; applying it to dwarf galaxies, with a 2.5 arcsecond matching radius and visual and spectroscopic interloper removal, produces the ten candidates. The Chomiuk and Wilcots relation is used to rule out individual or collective supernova remnants as the radio source.","core_discovery":"The central claim is that ten dwarf galaxies, each with stellar mass below about three billion solar masses and redshift at most 0.15, contain radio sources whose 3 GHz luminosity is too high relative to their total infrared luminosity to be explained by star formation. The excess is measured by $q = \\log(L_{\\rm IR}/(3.75\\times 10^{12}\\,L_{1.4\\,{\\rm GHz}}))$; sources with $q < 1.94$, two standard deviations below the peak of the calibration distribution, are classified as radio-excess AGNs. Statistical matching arguments predict fewer than one background interloper among the ten, and optical emission-line diagnostics place some of the hosts in AGN or composite regions, supporting physical association. The paper therefore presents these ten as strong intermediate-mass black hole candidates, five newly recognized as AGNs, and separately reports eight variable radio sources in dwarf galaxies found by comparing VLASS epochs to FIRST.","pith_inferences":["If the same $q<1.94$ threshold is applied to dwarfs with weak WISE detections, its validity rests on the infrared-radio scatter in dwarf galaxies being no larger than in the calibration sample; measuring the actual dwarf $q$ distribution would test whether any of the ten are ordinary low-$q$ star-forming outliers rather than AGNs.","Beyond the paper, the same selection on the completed three-epoch VLASS could multiply the dwarf AGN sample; the catalogs published here provide the selection function needed to estimate intermediate-mass black hole occupation fractions, which would bear directly on black hole seeding models.","The variable sources could also be followed up as tidal disruption event candidates, since the small black holes in dwarfs have strong tidal fields and the paper's luminosity arguments already exclude ordinary supernova remnants."],"forward_implications":["The ten dwarf galaxies become high-priority targets for X-ray, very long baseline interferometry, and optical follow-up; confirming any of them would add intermediate-mass black hole candidates near $10^5$-$10^6\\,M_\\odot$.","Five of the ten are identified as AGNs for the first time, showing that radio-excess selection catches dwarf AGNs that optical emission-line and mid-infrared color diagnostics miss.","The published catalogs of radio-excess AGNs and star-forming galaxies give the community a directly usable sample for measuring how the AGN fraction changes with galaxy mass and redshift.","Eight variable radio sources in dwarf galaxies, seen by comparing VLASS epochs with FIRST, imply that radio variability is common among these candidates and needs multi-band follow-up to distinguish AGN variability from transients such as tidal disruption events or radio-loud supernovae."],"supporting_citations":[{"why":"Supplies the q = q_peak - 2σ selection rule and the completeness/contamination compromise that justifies the threshold.","marker":"Delvecchio et al. (2021)"},{"why":"Established the IRRC parameter and the radio-excess approach in the local universe, giving the q baseline for star-forming galaxies.","marker":"Yun et al. (2001)"},{"why":"Provided the total-infrared version q_TIR = 2.64±0.26 against which the measured peak of 2.62 is checked.","marker":"Bell (2003)"},{"why":"Gives the relation L_TIR ≈ 8.33 L_25μm used to convert WISE W4 (22 μm) luminosity to total infrared luminosity.","marker":"Hao et al. (2011)"},{"why":"Shows 22 μm and 25 μm flux densities are approximately equal, completing the W4-to-TIR conversion.","marker":"Jarrett et al. (2013)"},{"why":"Supplies the K-correction formula and spectral-index treatment for converting 3 GHz VLASS fluxes to 1.4 GHz luminosities.","marker":"Novak et al. (2017)"},{"why":"The previous FIRST-based dwarf radio AGN search whose sample, interlopers, and variable-source lists are compared against throughout.","marker":"Reines et al. (2020)"},{"why":"Provides the superluminous-SNR luminosity relation and SNR luminosity function used to rule out supernova remnants as the radio source.","marker":"Chomiuk & Wilcots (2009)"}],"fun_headline_variants":["Ten dwarf galaxies reveal radio-excess black hole candidates","Radio survey finds 10 dwarf galaxies with black hole signs","Ten dwarf galaxies show radio-excess black hole activity","VLA survey reveals 10 dwarf galaxies with AGN candidates","Ten dwarf galaxies harbor radio-excess black hole candidates"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The ten candidates are selected with a $q<1.94$ cutoff calibrated on higher-mass galaxies with strong WISE detections, while several dwarf hosts have only upper limits on their infrared luminosity, so the selection assumes dwarf star-forming galaxies scatter around the infrared-radio relation no more than the calibration sample does.","fun_headline_variants_meta":{"raw":{"variants":["Ten dwarf galaxies reveal radio-excess black hole candidates","Radio survey finds 10 dwarf galaxies with black hole signs","Ten dwarf galaxies show radio-excess black hole activity","VLA survey reveals 10 dwarf galaxies with AGN candidates","Ten dwarf galaxies harbor radio-excess black hole candidates"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000624,"raw_usage":{"total_tokens":2953,"prompt_tokens":1072,"completion_tokens":1881,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":688,"completion_tokens_details":{"reasoning_tokens":1802}},"tokens_in":688,"tokens_out":1881,"duration_ms":13584,"temperature":1.0,"reasoning_tokens":1802,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-12T15:10:02.246915+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Measure the full $q$ distribution for a volume-limited sample of dwarf galaxies with no AGN signatures, using deep radio and infrared data; if more than about two percent of such dwarfs have $q<1.94$, the ten radio-excess AGN candidates could be ordinary low-$q$ star-forming galaxies rather than accreting black holes.","supporting_citations":[],"review_version":1}