{"id":"67932e09-b9c3-426a-9ee3-cb816968b1c5","arxiv_id":"1908.05229","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":5.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"Using SDSS spectra, the authors classify over 1800 sources from two new blazar candidate catalogs, find roughly 30% are confirmed blazars, identify the main contaminants, classify 11 Fermi uncertain blazars, and report 25 new BL Lac objects.","lead":"This paper tests two new catalogs of blazar candidates, built from infrared colors, against optical spectra from the Sloan Digital Sky Survey. It finds that only about 30 percent of the candidates are confirmed blazars, with quasars and galaxies making up most of the rest.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Lower-limit claim depends on unobserved sources not being blazar-rich; missing spectra correlate with faintness.","rationale":"The reader's weakest assumption correctly identifies the selection effect on spectroscopic completeness as the crux: the observed SDSS subsamples are not random subsets, and the paper does not establish that unobserved sources have blazar fractions at least as high as observed ones. The manuscript itself flags this in Section 4.3 (missing KDEBLLACS sources are fainter) and Section 5 (contamination estimates as upper limits), but the Summary still phrases the result as 'at least ~30% of each catalog,' which is stronger than the presented evidence. The concern is internally grounded: Figure 9 shows a magnitude offset for KDEBLLACS, and the classification results already show a large galaxy/LLAGN fraction among the observed faint sources. This is a correctness risk for the central claim, not mere disagreement with consensus. My concrete test would settle it by checking whether blazar fraction varies with magnitude, directly probing whether the fainter unobserved population is likely to dilute or enhance the reported fraction. I found no independent evidence in the paper that mitigates this: the color-color tests in Section 4.3 address color selection, not magnitude-dependent incompleteness, and there are no error bars on the reported fractions. The verdict CONDITIONAL matches the reader's assessment: the paper should be accepted only if the authors soften the lower-limit statements or supply a completeness-corrected estimate, since the spectral classifications and new identifications remain valuable regardless.","tokens_in":13430,"tokens_out":1639,"duration_ms":14662,"concrete_test":"Recompute the blazar fraction lower limit using only sources brighter than the spectroscopic completeness limit of SDSS DR15 (e.g., r < 17.77 for the main galaxy sample), and separately compute the blazar fraction as a function of r-band magnitude for the observed subsamples. If the magnitude-binned blazar fraction is flat or decreasing toward faint magnitudes, the 'at least ~30%' claim is supported; if it rises at faint magnitudes, the lower limit is not rigorous and should be replaced with an explicitly subsample-qualified statement or a completeness-corrected estimate.","verdict_should_be":"CONDITIONAL","load_bearing_attack":"The paper's headline result, 'at least ~30% of each catalog is composed by confirmed blazars,' is established only for the SDSS-spectroscopically observed subsamples: 1424 of 9541 WIBRaLS2 sources (15%) and 406 of 5580 KDEBLLACS sources (7%). Sections 4.3 and 5 correctly state that contamination estimates are upper limits, but the Summary also asserts lower limits on blazar efficiency without a rigorous argument that the unobserved sources are not blazar-rich. Figure 9 shows that KDEBLLACS sources lacking spectra are systematically fainter than those with spectra; the paper itself notes that KDEBLLACS candidates are intrinsically fainter in WISE bands (Section 4.3, citing D'Abrusco et al. 2019). Fainter objects are more likely to be host-galaxy-dominated or normal galaxies, and the spectroscopically classified KDEBLLACS sample already contains 155 normal galaxies and 47 LLAGNs versus only 75 BZBs. If the fainter unobserved population is even more galaxy-dominated, the true blazar fraction could fall below 30%, invalidating the 'at least' claim. The same selection concern applies to WIBRaLS2, though Figure 9 shows similar magnitude distributions there; the paper's color-based selection-effect test in Section 4.3 does not directly probe the magnitude-dependent spectroscopic completeness that Figure 9 reveals. Thus the lower-limit inference is the load-bearing weak point.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper characterizes two WISE-selected blazar candidate catalogs, WIBRaLS2 and KDEBLLACS, using optical spectra from SDSS DR15. The authors cross-match catalog sources to SDSS spectra, visually classify the spectra into blazars, QSOs, galaxies, stars, and other classes, and report the fraction of confirmed blazars in each catalog. They find that the spectroscopically observed subsamples contain roughly 30% confirmed blazars, with QSOs being the main contaminants in WIBRaLS2 and normal galaxies in KDEBLLACS. They also classify 11 Fermi 4FGL BCUs and report 25 new BL Lac objects. The paper's headline claim is that 'at least ~30%' of each full catalog consists of confirmed blazars, presented as a lower limit on selection efficiency.","tokens_in":13743,"tokens_out":3853,"duration_ms":37213,"significance":"If the result holds, this is a useful benchmark for the community, because WIBRaLS2 and KDEBLLACS are widely used for Fermi counterpart identification and blazar population studies. The analysis is independent of the catalogs' construction: it relies on external SDSS DR15 spectra and the Roma-BZCAT catalog as benchmarks, so the core classification test is not circular. The identification of 11 BCU counterparts and 25 new BZBs is a concrete contribution. The main weakness is that the lower-limit interpretation of the blazar fraction is not rigorously supported by the data, which means the headline quantitative claim needs revision or additional analysis. The observed-subample fractions are still valuable, but they should be reported as such unless the selection-effect issue is addressed.","major_comments":[{"comment":"The abstract and summary claim that 'at least ~30%' of each catalog is composed of confirmed blazars, but this lower-limit interpretation is not supported by the analysis. Only 1424 of 9541 WIBRaLS2 sources (~15%) and 406 of 5580 KDEBLLACS sources (~7%) have SDSS spectra, and the paper's own Figure 9 shows that KDEBLLACS sources lacking spectra are systematically fainter than those with spectra. Section 4.3 also notes that KDEBLLACS candidates are intrinsically fainter in WISE bands. Fainter sources are more likely to be host-galaxy-dominated or normal galaxies, and the spectroscopically classified KDEBLLACS sample already contains 155 normal galaxies versus only 75 BZBs. The statement that contamination estimates are 'upper limits' does not imply that the blazar efficiency is a lower limit; a lower bound would require an explicit argument that the unobserved sources are not more blazar-rich than the observed ones, which is absent. The reported fractions (34.6%, 27.7%, and 42.5% for WBZB, WBZQ, and WMIXED in WIBRaLS2, and ~30% for KDEBLLACS) should be presented as observed-subample fractions, or the lower-limit claim needs a quantitative treatment of the magnitude-dependent incompleteness.","section":"Section 5 (first two bullets) and Section 4.3, Figure 9"},{"comment":"The classification fractions are reported without statistical uncertainties. For example, '17.6% of WBZBs show a featureless optical spectrum' and '26.1% of WBZQs are confirmed BZQs' are point estimates based on subsamples of 471 and 833 objects, respectively; without binomial or Poisson confidence intervals, the reader cannot assess the significance of differences such as the 30.8% versus 26.1% BZQ fractions between WMIXED and WBZQ sources. Adding uncertainties would strengthen the quantitative claims about the dominant contaminant classes and the relative purity of the color-selected subregions.","section":"Sections 4.1 and 4.2"},{"comment":"The color-based selection-effect test concludes that 'we can rule out the existence of significant selection effects due to the colors of the optical counterparts,' but the analysis only compares isodensity contours in optical color-color space; it does not directly probe the joint color-magnitude selection function that determines SDSS spectroscopic targeting. Given that Figure 9 demonstrates a magnitude-dependent incompleteness, a test that does not incorporate magnitude cannot fully rule out selection effects that correlate with color and magnitude simultaneously. At minimum, the conclusion should be restricted to 'no strong color-only selection effect' rather than the current broader statement.","section":"Section 4.3, Figure 8"}],"minor_comments":[{"comment":"The caption lists 'W1, W1, W3 and W4' but should read 'W1, W2, W3 and W4'.","section":"Figure 1 caption"},{"comment":"The word 'splitted' should be 'split'.","section":"Section 4, first paragraph"},{"comment":"The abstract reports 'at least ~30%' while the summary gives a weighted lower limit of ~31% for WIBRaLS2 and ~30% for KDEBLLACS; these numbers should be made consistent and, if the lower-limit language is removed as suggested above, the abstract and summary should be reworded to describe the observed-subample fractions.","section":"Abstract and Section 5"},{"comment":"The text says 'This sample corresponds to 7% of the whole KDEBLLACS' after stating the final sample is 406 sources; 406/5580 is approximately 7.3%, so the rounding is acceptable but could be clarified.","section":"Section 2.2"},{"comment":"The captions for Figures A.1 and A.2 state that 'all sources present featureless spectra with color index u− r < 1.4,' but Table 2 lists one source with u− r = 1.38 and another with 1.32; the criterion should be described as 'u− r ≲ 1.4' or the threshold should be clarified to avoid an apparent inconsistency.","section":"Appendix A caption"}],"recommendation":"major_revision","confidential_remarks":"The paper is within the scope of A&A and the observational characterization is useful. The main issue is the unsupported lower-limit interpretation of the blazar fraction; this is fixable by rephrasing the claims to refer to the observed SDSS subsample or by adding a quantitative treatment of the magnitude-dependent incompleteness. I do not see circularity as a concern, since the spectral classification relies on external SDSS and Roma-BZCAT data. The lack of uncertainties on the classification fractions is secondary but should be addressed in revision. I recommend major revision rather than rejection because the core dataset and classification effort are sound."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The obvious but necessary job: take the two new WISE-selected blazar candidate catalogs, WIBRaLS2 and KDEBLLACS, cross-match them to SDSS DR15 spectra, and classify the sources. This is the first optical characterization of either catalog, and it yields real new results: 25 previously unknown BL Lac objects and spectral classifications for 11 4FGL BCUs. Using external SDSS spectra and Roma-BZCAT as a benchmark means the test is not circular with respect to the catalog selection. The KDE contamination maps are a nice visual summary that will help people plan follow-up.\n\nThe soft spot is the headline claim. The abstract and summary say 'at least ~30% of each catalog' is composed of confirmed blazars, but the observed samples are only 15% of WIBRaLS2 and 7% of KDEBLLACS. Figure 9 shows that KDEBLLACS sources without spectra are systematically fainter than those with spectra, and the paper itself notes in Section 4.3 that KDEBLLACS candidates are intrinsically fainter in WISE bands. The classified KDEBLLACS sample already contains more normal galaxies and LLAGNs than BZBs, so if the fainter unobserved tail is even more galaxy-dominated, the true blazar fraction could fall below 30%. The authors partially acknowledge this when they call the contamination estimates upper limits in Section 5, but then turn around and call the efficiencies lower limits. That is not rigorous. For WIBRaLS2 the magnitude distributions are similar, so the concern is less severe, but the same logic applies to the roughly 52% without spectra.\n\nThe new BZBs and BCU classifications are independent of the lower-limit claim and stand on their own. Minor issues: no uncertainties on the derived fractions, and the spectral classification is not independently double-checked, but the criteria are standard. The fix is wording, not a new analysis: either soften the lower-limit language to describe the observed subsample, or provide a quantitative argument that the unobserved sources are not blazar-rich.\n\nI would send this to a referee. It is a useful catalog characterization that the blazar and gamma-ray follow-up community will cite. The central claim needs a caveat, but the work is worth publishing after a moderate revision.","headline":"A solid spectroscopic characterization of two new blazar catalogs; the headline efficiency numbers need a caveat about the fainter unobserved sources.","tokens_in":14263,"tokens_out":2431,"would_cite":true,"duration_ms":23349,"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":"Optical spectroscopy of 1830 WISE-selected candidates shows that only about 30% of each catalog is made of confirmed blazars, with quasars and galaxies as the main contaminants.","keywords":["WIBRaLS2","KDEBLLACS","blazar candidates","WISE infrared colors","SDSS optical spectra","BL Lac objects","gamma-ray source counterparts","active galactic nuclei"],"falsifier":"Take a random or complete sample of the WIBRaLS2 and KDEBLLACS sources in the SDSS footprint that currently lack spectra, obtain deep optical spectra for them, and compare the confirmed-blazar fraction with the reported ~30% lower limit; if the faint missing sources are predominantly normal galaxies, the true blazar fraction will fall below ~30%.","tokens_in":13286,"feed_emoji":"🔭","tokens_out":5626,"duration_ms":49022,"temperature":0.7,"pith_summary":"This paper asks how trustworthy two newly released catalogs of blazar candidates—objects chosen by their infrared colors in WISE data and by radio loudness—really are. By examining optical spectra from SDSS DR15 for 1830 catalog sources, the authors find that at least roughly 30% of each catalog consists of spectroscopically confirmed blazars. The main impostors differ: quasars dominate the contamination of WIBRaLS2 (~58%) while normal galaxies dominate KDEBLLACS (~38.2%). The same spectra let the authors pin down the nature of 11 previously uncertain Fermi-LAT blazar candidates and identify 25 new BL Lac objects. A sympathetic reader would care because these catalogs are used to find counterparts for the large share of Fermi gamma-ray sources that remain unassociated.","feed_headline":"Spectra show WISE blazar catalogs are only ~30% blazars","feed_subtitle":"1830 SDSS optical spectra reveal quasars and galaxies as the main contaminants of both catalogs.","key_machinery":"The central machinery is the WISE Blazar Strip and its three-dimensional WISE color locus: WIBRaLS2 selects sources whose W1-W2 and W2-W3 colors match confirmed Fermi blazars and that are radio-loud, while KDEBLLACS applies a kernel density estimate to the two-dimensional WISE color diagram and keeps sources within the 90% isodensity contour of known BL Lacs, also requiring radio loudness. To test these selections, the authors classify SDSS DR15 optical spectra using established diagnostics: featureless blue continuum for BL Lacs, broad emission lines plus radio luminosity for quasars, the Ca II break depression parameter $C_B$ with threshold 0.25 to separate BL Lac-host galaxies from normal galaxies, and BPT line ratios to separate low-luminosity AGNs from star-forming galaxies. Kernel density estimates applied to contaminants and to 4FGL gamma-ray blazars in the WISE color space localize the contamination to the edges of the selection region.","core_discovery":"The paper reports that WIBRaLS2 and KDEBLLACS, though built from WISE mid-infrared colors similar to known gamma-ray blazars, are not highly pure: for the SDSS-observed subsample, the confirmed-blazar fraction is about 31% for WIBRaLS2 (34.6% of candidate BZBs, 27.7% of candidate BZQs, 42.5% of mixed-color candidates) and about 30% for KDEBLLACS; the authors treat these as lower limits for the whole catalogs. Quasars are the main contaminant of WIBRaLS2 (~58%) and normal galaxies of KDEBLLACS (~38.2%), with contaminants concentrated toward the edges of the infrared color-color diagram. In addition, optical spectra yield classifications for 11 4FGL blazar candidates of uncertain type and reveal 25 new BL Lac objects not previously cataloged.","pith_inferences":["If the fainter unobserved KDEBLLACS sources are mostly galaxies, the catalog's true blazar fraction could fall below 30%; testing this requires spectroscopy of a magnitude-limited sample beyond SDSS's current depth.","Some of the WIBRaLS2 QSO contaminants may actually be flat-spectrum radio quasars, which are a subclass of blazars; if so, the effective blazar content could be higher than 30%, and radio spectral indices of those contaminants would discriminate.","The $u-r<1.4$ criterion used to vet the new BL Lacs is stated to fail at redshifts above 0.5, so a few of the 25 new objects at higher redshift could be misclassified; redshift-resolved spectroscopy would settle their nature.","The edge-concentration pattern of contaminants suggests a reusable test: applying the same kernel-density contamination mapping to future color-selected AGN catalogs could reveal where their purity drops before committing to large follow-up campaigns."],"forward_implications":["WIBRaLS2 and KDEBLLACS can be used to propose counterparts for unassociated Fermi-LAT sources, but with a known purity floor around 30%; candidates lying in the more contaminated regions of the WISE color-color diagram should be treated with lower confidence.","Contaminants are not random: quasars cluster in the WBZQ region of WIBRaLS2 and galaxies and quasars sit at the edges of the KDEBLLACS selection, so targeted cuts could improve the purity of future catalog releases.","The 25 newly identified BL Lac objects expand the spectroscopic census of this class and can serve as training samples for future infrared- or radio-based selection techniques.","The 11 classified 4FGL blazar candidates of uncertain type reduce the uncertain fraction of the gamma-ray catalog, sharpening population studies of the gamma-ray sky."],"supporting_citations":[{"why":"Builds and presents the two catalogs being tested, including the WISE color criteria, radio-loudness parameters, and the kernel density selection used for KDEBLLACS.","marker":"D’Abrusco et al. 2019"},{"why":"Provides SDSS DR15, the source of all optical spectra used for the classification.","marker":"Aguado et al. 2019"},{"why":"Roma-BZCAT supplies the spectroscopically confirmed blazar classifications that anchor the catalog purity estimates.","marker":"Massaro et al. 2015a"},{"why":"Defines the 4FGL catalog, including the blazar-candidates-of-uncertain-type class whose members are classified here and the gamma-ray blazars used in the kernel density comparisons.","marker":"Fermi-LAT Collaboration et al. 2019"},{"why":"Defines the Ca II break parameter and the $u-r$ color criterion used to separate BL Lacs from galaxies and to vet the new BL Lac candidates.","marker":"Massaro et al. 2012a"},{"why":"Supplies the $C_B$ threshold of 0.25 used to classify a source as a BL Lac-host galaxy rather than a normal galaxy.","marker":"Stocke et al. 1991"},{"why":"Introduces the BPT line-ratio diagram used to separate low-luminosity AGNs from star-forming galaxies.","marker":"Baldwin et al. 1981"},{"why":"Provides the theoretical boundary used in the BPT diagram classification in this work.","marker":"Kewley et al. 2006"}],"fun_headline_variants":["WISE blazar catalogs: only ~30% are true blazars","Spectra show WISE blazar catalogs ~30% pure","WISE blazar lists: one-third blazars, two-thirds not","Optical check: WISE blazar catalogs ~30% genuine","Quasars and galaxies dominate WISE blazar catalogs"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The results treat the SDSS-observed subsample as giving a lower limit for the whole catalog, which requires that the unobserved sources (about half of WIBRaLS2 and three-quarters of KDEBLLACS in the SDSS footprint, mostly fainter) are not more galaxy-dominated than the observed ones.","fun_headline_variants_meta":{"raw":{"variants":["WISE blazar catalogs: only ~30% are true blazars","Spectra show WISE blazar catalogs ~30% pure","WISE blazar lists: one-third blazars, two-thirds not","Optical check: WISE blazar catalogs ~30% genuine","Quasars and galaxies dominate WISE blazar catalogs"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000489,"raw_usage":{"total_tokens":2493,"prompt_tokens":1115,"completion_tokens":1378,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":731,"completion_tokens_details":{"reasoning_tokens":1282}},"tokens_in":731,"tokens_out":1378,"duration_ms":12348,"temperature":1.0,"reasoning_tokens":1282,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T13:19:42.195130+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Take a random or complete sample of the WIBRaLS2 and KDEBLLACS sources in the SDSS footprint that currently lack spectra, obtain deep optical spectra for them, and compare the confirmed-blazar fraction with the reported ~30% lower limit; if the faint missing sources are predominantly normal galaxies, the true blazar fraction will fall below ~30%.","supporting_citations":[{"cited_title":"2019, The Astrophysical Journal Supplement Series, 240, 23","cited_arxiv_id":null,"evidence_quote":"Provides SDSS DR15, the source of all optical spectra used for the classification."},{"cited_title":"T., Morris, S","cited_arxiv_id":null,"evidence_quote":"Supplies the $C_B$ threshold of 0.25 used to classify a source as a BL Lac-host galaxy rather than a normal galaxy."},{"cited_title":"A., Phillips, M","cited_arxiv_id":null,"evidence_quote":"Introduces the BPT line-ratio diagram used to separate low-luminosity AGNs from star-forming galaxies."},{"cited_title":"J., Groves, B., Kauffmann, G., & Heckman, T","cited_arxiv_id":null,"evidence_quote":"Provides the theoretical boundary used in the BPT diagram classification in this work."}],"review_version":1}