{"id":"64c85c21-dd31-4d1c-9de2-9355be40be7e","arxiv_id":"2412.01887","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":5,"one_line_summary":"A new photometric selection recovers 248 Little Red Dots across JWST fields, increasing the known sample by about 1.7x, and reveals warm dust emission in one low-redshift LRD.","lead":"Astronomers present a new color-based method to find 'Little Red Dots', compact red galaxies from the early universe, and use it to identify 248 such objects, about 70% more than earlier searches. One of these galaxies shows the first clear sign of warm dust glowing in infrared, which may help explain what these mysterious objects are.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Selection purity is untested: no redshift lower bound and no spectroscopy for the 99 new LRDs, so the ×1.7 density increase may be inflated by low-z compact quiescent galaxies rather than real LRDs.","rationale":"The reader's weakest assumption concerns completeness: the five empirical LRD models may not span the true LRD locus, so real LRDs could be missed and the ×1.7 gain could be a calibration artifact. My concern is the mirror-image issue of purity: the selection may also include objects that are not LRDs, inflating the census. Both concerns attack the external validity of the photometric selection, but they are distinct. I regard contamination as the more load-bearing risk because the paper itself demonstrates high recovery of previous samples (93%), making incompleteness less likely than over-inclusion; moreover, the absence of any redshift lower bound and of spectroscopy for the new candidates makes contamination directly testable. This concern does not change the reader's CONDITIONAL verdict; it strengthens it. The paper is otherwise transparent and careful: the selection is precisely defined, the comparison to previous samples is useful, and the MIPS24 deblending appendix (§A) is a genuine attempt to address companion contamination. The warm-dust claim for JADES-191056 is also somewhat fragile, given the unresolved PACS photometry and proximity of the companion, but the census claim is more central to the paper's stated goal of a more complete LRD population. A conditional verdict requiring spectroscopic or quantitative purity assessment is therefore appropriate.","tokens_in":37829,"tokens_out":8960,"duration_ms":101833,"concrete_test":"Cross-match all 248 candidates against public spectroscopic catalogs in the JADES/GOODS-S, COSMOS, and UDS fields (e.g., MUSE, 3D-HST, zCOSMOS, DEIMOS, VANDELS). For the ~99 objects not in KOV24/KOI24, tabulate spectroscopic redshifts and spectral classifications; require that >80% have z>3 and show LRD diagnostic features (broad Balmer lines, compact blue+red V-shaped SED). If >20% are z<2.5 galaxies without such features, the ×1.7 census gain is spurious. As a complementary check, run the same color and compactness cuts on a mass-complete sample of spectroscopically confirmed z<3 compact quiescent galaxies to estimate the expected contamination fraction.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central census claim (§3.3, §6) rests on the 248 selected objects being predominantly genuine LRDs at 3<z<9. The color cuts (i)–(iv) in §3.1 were calibrated using model tracks only over z=3–9 (Figure 1), but no photometric-redshift or spectroscopic lower bound is applied to the sample; the only additional cut is compactness (§3.2). At z≲3, F115W−F200W probes rest-frame optical rather than blue rest-frame UV, so compact quiescent or dusty star-forming galaxies with a modestly rising red SED can satisfy F200W−F444W > F115W−F200W + 0.25 and enter the selection region. Redshifts are then estimated with EAZY using an LRD template and refined with Prospector-AGN (§2.2, §4.1); a low-z compact red galaxy can plausibly be fit at higher z, so the quoted 3<z<9 distribution does not by itself demonstrate purity. The 99 objects not in KOV24/KOI24 are preferentially faint (F444W≈26.2) and blue in F277W−F444W (≈0.8), and no spectroscopic follow-up is presented for them. If a substantial fraction are low-z interlopers, the ×1.7 number-density increase is a selection artifact rather than evidence that previous censuses were underestimated.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper presents a new photometric selection for Little Red Dots (LRDs) based on NIRCam colors (F200W−F444W>1, F200W−F444W > F115W−F200W+0.25, F115W−F200W>−0.5, F444W<27) plus a compactness criterion, applied to 263 arcmin^2 of JADES, PRIMER-COSMOS, and PRIMER-UDS fields with MIRI coverage. The selection yields 248 LRDs, about 1.7 times more than previous samples in the same fields, with the new objects preferentially fainter and bluer. The paper uses MIRI photometry to derive rest-frame UV-to-NIR and near-to-mid IR colors, concluding that most LRDs show some dust emission at 3 μm but much less than obscured QSOs. A detailed case study of JADES-191056 at z=3.1386 combines MIRI, Spitzer/MIPS24, and Herschel/PACS data to claim the first direct evidence of warm dust emission (T~50–100 K) in an LRD. The SED is fit with two independent codes (Prospector-AGN and Synthesizer-AGN), but galaxy- and AGN-dominated scenarios remain degenerate.","tokens_in":38226,"tokens_out":2326,"duration_ms":27699,"significance":"If the selection is pure and complete as claimed, the paper's central result—that previous LRD censuses underestimated the population by a factor of ~1.7—is an important step toward a complete census of these objects at 3<z<9. The paper also provides a valuable MIRI-based color framework for distinguishing dust-dominated from dust-poor LRDs, and the MIPS24/PSF decomposition for JADES-191056 is a careful treatment of a blended source. The methodological strengths include the use of two independent SED-fitting codes, explicit comparison with previous samples (KOV24, KOI24), and public MIRI data. The main caveat is that the selection thresholds are calibrated on a small set of model tracks, and the redshift distribution is derived from photometric redshifts without an explicit purity test, so the magnitude of the density increase is not yet fully secured.","major_comments":[{"comment":"The selection thresholds (i)–(iv) are explicitly chosen to enclose the color-redshift tracks of five LRD models and to exclude a star-forming and a type-1 QSO track. This is a calibration loop: the selected sample will trivially contain objects with colors similar to the models. The claim that the sample is more complete (×1.7) than previous selections requires that those five models span the full LRD locus, but the paper does not provide a quantitative test of how the completeness fraction varies across the color-magnitude diagram, nor does it quantify the fraction of the 248 objects that lie in regions of the diagram not sampled by the model tracks. I recommend adding a completeness estimate computed from simulated sources drawn from a broader distribution of SEDs (including low-z interlopers) and evaluating the selection's purity against spectroscopic redshift samples in these fields.","section":"§3.1, Figure 1"},{"comment":"The number-density comparison (0.93 arcmin^-2, ×1.7 relative to KOV24/KOI24) rests on the assumption that the 248 selected objects are predominantly genuine LRDs at 3<z<9. However, the selection applies no redshift lower bound, and the only non-color cut is the compactness ratio (FF444W(0.5\")/FF444W(0.2\")<1.5). At z≲3, F115W−F200W probes rest-frame optical rather than UV, and compact quiescent or dusty star-forming galaxies can plausibly satisfy the diagonal color cut. The paper's redshift distribution is based on EAZY with an LRD template and Prospector-AGN fits; a low-z compact red galaxy can be fit at higher redshift, so the plotted 3<z<9 histogram does not demonstrate purity. The 99 objects not in previous samples are preferentially faint (median F444W≈26.2) and blue (median F277W−F444W≈0.8), making them especially vulnerable to contamination. I request a quantitative contamination assessment: e.g., how many selected objects have photometric redshifts <3 from an independent code, how many have spectroscopic redshifts (including MUSE in GOODS-S), and what fraction of the new faint/blue subset is expected to be low-z interlopers based on the non-LRD models shown in Figure 1.","section":"§3.3, §4.1, §6"},{"comment":"The redshift distribution shows a median z=5.5 with a peak at z=5 and secondary peak at z=7, but the text does not state whether any redshift cut is applied to the sample before computing number densities. If objects with z<3 are included in the 248 count, the ×1.7 density increase is inflated by non-LRDs. The paper should either apply a hard photometric-redshift lower bound (and show how many objects are removed) or explicitly justify why no lower bound is needed despite the selection's reduced ability to distinguish LRDs from red galaxies at low z. This is load-bearing for the central census claim.","section":"§4.1, Figure 6 (bottom)"},{"comment":"For JADES-191056, the MIPS24 flux is derived via PSF fitting with position priors from MIRI, yielding fν=35.7±3.8 μJy for the LRD and 12.3±1.7 μJy for the companion. However, the PACS100 and PACS160 fluxes are measured in circular apertures without deconvolution, and the text states the companion contributes ~30% based on the MIPS24 analysis. The warm-dust claim (steep rise at λ_rest>6 μm, peak ~40 μm, T=50–100 K) depends on the PACS photometry. Please quantify how the conclusion changes if the companion contributes more than 30% in the PACS bands (e.g., if the companion has a redder mid-IR SED), or provide an upper-limit analysis that excludes PACS160 given its low SNR.","section":"§5, Figure 11, Appendix A"}],"minor_comments":[{"comment":"There is a typo: 'stidues' should be 'studies'.","section":"§5 (first paragraph)"},{"comment":"The center panel caption begins 'Center: Center:' with a duplicated word.","section":"Figure 7 caption"},{"comment":"The text says PRIMER has 'the same WCF3 bands'—should be 'WFC3 bands'.","section":"§2.1"},{"comment":"The caption describes 'KOK24' in one place ('KOK24 (purple)') while the text and legend use KOV24; please standardize the abbreviation.","section":"Figure 3 caption"},{"comment":"The statement that F1800W detections are concentrated at z≲5 with median z=4.7 is based on only 17 objects; the text should note the small-number statistics when interpreting this subset.","section":"§4.1"}],"recommendation":"major_revision","confidential_remarks":"The paper is a solid observational study with a clear and useful selection method, but the central census claim (×1.7 density increase) hinges on sample purity, which is not yet demonstrated. The calibration loop between model tracks and thresholds is a real concern, and the lack of an independent purity test (spectroscopic or photometric-redshift based) makes the headline number somewhat fragile. The JADES-191056 warm-dust detection is interesting but also depends on the uncertain PACS photometry in a blended system. I would be supportive of publication after the authors add a quantitative contamination estimate and test the robustness of the density increase to reasonable low-z interloper fractions."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Let me give you my take on Barro et al. The headline result is a new LRD color–color selection that recovers 93% of prior samples and adds ~40% new objects, raising the number density by 1.7×. That is worth knowing about regardless of whether the exact factor holds up. The paper also presents JADES-191056, a z=3.1386 LRD with a steep mid-IR rise and a ~40 µm peak, which they argue is the first direct warm-dust detection in an LRD.\n\nThe selection itself is a genuine improvement: using F200W–F444W and F115W–F200W gives a longer baseline and reduces the emission-line spikes that afflict F277W-based cuts. The overlap comparison between KOV24 and KOI24 (only ~50%) is a useful community result. The SED work on the two low-z objects, especially the MIPS24 deblending with MIRI priors, is careful.\n\nThe soft spot is the census claim. The 99 new objects are faint (F444W~26.2) and blue in F277W–F444W, and none have spectroscopy. The selection thresholds were tuned to five model tracks, and the photometric redshifts are estimated with an LRD template, so the quoted 3<z<9 distribution is not an independent check. Compact quiescent or dusty star-forming galaxies at z~2–3 can plausibly enter that color space. The paper applies a compactness cut, but that does not remove a genuinely compact low-z galaxy. So the ×1.7 may be inflated by interlopers. The paper does not quantitatively estimate contamination—no control sample, no application of the selection to a low-z redshift-selected sample, no use of existing spectroscopy in the fields. That is the main thing a referee should push on.\n\nThe warm-dust detection is more solid as a detection but degenerate in interpretation. The MIPS24 flux is deblended, the PACS points are low-SNR, and the paper honestly says the SED can be fit either with ISM dust or a warm AGN torus. So 'first direct evidence' is okay if that is understood as 'first direct photometric detection of the mid-IR rise,' not a determination of the dust geometry.\n\nI think the paper deserves a serious referee. The selection is a useful consolidation even if the density factor changes, and the JADES-191056 analysis will be cited. But the referee should ask for a contamination estimate: apply the selection to known low-z galaxies in the same fields, or present the photometric-redshift quality and show that the new objects have multi-band SEDs inconsistent with z<3 templates. If the authors can do that, the result becomes much stronger.","headline":"Useful new LRD selection that recovers almost all previous samples and adds 40% more candidates, but the ×1.7 census gain rests on unverified purity and needs a contamination test; the warm-dust source is a careful, if degenerate, case study.","tokens_in":38803,"tokens_out":3027,"would_cite":true,"duration_ms":33141,"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":"A new photometric selection using a diagonal NIRCam color cut recovers 93% of known Little Red Dots and adds 40% more, raising their number density by a factor of 1.7, while MIRI data reveal warm dust in a z=3.1386 LRD.","keywords":["Little Red Dots","photometric selection","JWST MIRI","number density","warm dust","AGN","high-redshift galaxies","dust emission"],"falsifier":"A MIRI-selected, spectroscopically confirmed sample of LRDs at z≈3–5 with F200W−F444W<1 mag, or a demonstration that the five model tracks misplace the observed LRD locus in Figure 1, would falsify the completeness claim. Alternatively, a complete F770W-selected census in the same fields yielding a surface density consistent with earlier samples (~0.5 arcmin⁻²) would show the ×1.7 gain is a selection artifact.","tokens_in":1990,"feed_emoji":"🔴","tokens_out":2813,"duration_ms":64200,"temperature":0.7,"pith_summary":"This paper introduces a photometric selection designed to capture the full range of colors and redshifts of 'Little Red Dots' (LRDs), compact red sources at z≈3–9 that are likely dusty starbursts or reddened AGN. Applied to 263 arcmin² of JWST NIRCam+MIRI imaging, the method finds 248 LRDs to F444W<27 mag, a factor of 1.7 higher number density than previous samples, implying earlier censuses underestimated the population. Using MIRI to extend the SED into the rest-frame near- and mid-infrared, the paper shows most LRDs are detected at 7.7 µm and that their rest-frame [1–3 µm] colors indicate diverse dust emission, with about half showing substantial dust fractions. It also reports a spectroscopically confirmed LRD at z=3.1386 whose SED rises steeply beyond 6 µm and peaks near 40 µm, interpreted as the first direct evidence of warm (50–100 K) dust emission in an LRD. A sympathetic reader would care because the census underpins estimates of early black hole growth and dust production.","feed_headline":"New selection nearly doubles census of little red dots","feed_subtitle":"A broader NIRCam color cut over 263 arcmin² recovers 248 LRDs and one z=3.14 source with 50–100 K dust.","key_machinery":"The selection is carried by a single NIRCam color-color diagram with a diagonal threshold: F200W−F444W > F115W−F200W + 0.25, along with F200W−F444W>1, F115W−F200W>−0.5, F444W<27. The pivot F200W traces rest-frame UV at z>4.5 and avoids the [O III]/Hα spikes that contaminate F277W; the diagonal cut runs parallel to the locus of massive dusty galaxies, separating LRDs from red extended galaxies while keeping the blue UV emission that defines them. A compactness criterion (F444W flux ratio 0.5\"/0.2\" <1.5) removes resolved interlopers. The color-redshift tracks of five LRD models spanning [0.25–1 µm]~2–5 mag define the selection region and are the device that produces the claimed completeness gain.","core_discovery":"The central claim is that a color-color selection using F200W−F444W>1, (F200W−F444W)>(F115W−F200W)+0.25, F115W−F200W>−0.5, F444W<27, plus a compactness cut, recovers 93% of previously known LRDs and adds ~40% new ones, raising the surface density to 0.93 arcmin⁻² and the number density by ×1.7 relative to KOV24 and KOI24. The new objects are preferentially fainter (median F444W=26.2) and bluer (F277W−F444W~0.8), many at z<5.5 where previous cuts were confused by [O III] emission. The paper's second claim is that MIRI data trace dust: F1800W-detected LRDs have median [1–3 µm]=1.5 mag, most have dust fractions up to f$_{3\\mu m}$~0.7–0.8, and one object, JADES-191056 at z=3.1386, shows a steep IR rise and peak at ~40 µm consistent with warm dust at T=50–100 K, the first direct evidence of such emission in an LRD.","pith_inferences":["The correlation between bluer UV-to-NIR and redder [1–3 µm] colors suggests a sequence where the bluest LRDs are transitional to unobscured type-1 QSOs; the paper hints at this but does not claim it as a unified evolutionary track.","If warm dust at 50–100 K is common, the ALMA non-detections of LRDs would imply the dust is compact and warm rather than extended and cold, which would favor AGN-torus or nuclear starburst geometries over galaxy-wide cold dust.","Deeper F1800W imaging by ~1 mag, as the paper notes, would turn most current upper limits into detections or tight bounds, directly testing whether the majority of LRDs have dust fractions as high as 0.7–0.8.","The same color-color method could be applied to wider JWST fields without MIRI, since the selection itself uses only NIRCam; MIRI is needed only for the dust-emission analysis, so the completeness gain may extend to the full NIRCam survey area."],"forward_implications":["If the selection is complete, previous LRD volume densities of ~10⁻⁵ Mpc⁻³ are underestimated by a factor ~1.7, strengthening the case that these objects are a major channel for early black hole growth.","The ~40% of new LRDs at z<5.5 with blue F277W−F444W colors imply that emission-line-boosted photometry had been hiding a substantial low-redshift LRD population.","The median [1–3 µm]=1.5 mag and dust fractions up to ~0.8 imply that dust emission, not a flat stellar continuum, dominates the 3 µm rest-frame light in most F1800W-detected LRDs.","The z=3.1386 LRD's warm-dust peak at ~40 µm makes low-redshift LRDs the best targets for future MIRI spectroscopy to identify PAH or silicate features and settle the ISM-versus-torus origin of the dust."],"supporting_citations":[{"why":"Supplies the five LRD SED models, the F277W-based selection, and the Prospector-AGN/Synthesizer-AGN codes used for rest-frame colors and dust modeling.","marker":"Pérez-González et al. (2024)"},{"why":"Provides the comparison sample KOI24, the redshift-dependent band selection, and the compactness criterion adopted here.","marker":"Kocevski et al. (2024)"},{"why":"Provides the comparison sample KOV24 and the red1/red2 color criteria that partially overlap with the new selection.","marker":"Kokorev et al. (2024a)"},{"why":"Provides the reddest LRD model used in the track calibration and the prior detailed study of UDS-40579.","marker":"Wang et al. (2024a)"},{"why":"Provides the large-area comparison number density and the IR SED stack used as context for the dust-emission interpretation.","marker":"Akins et al. (2024)"},{"why":"Provides MIRI photometric procedures and the hot-dust fraction interpretation for the [1–3 µm] color models.","marker":"Leung et al. (2024)"}],"fun_headline_variants":["Little red dot census jumps 70% with new color selection","First warm dust emission seen in a little red dot","New photometric cut uncovers 1.7× more little red dots","JWST survey: 248 little red dots, one with 50–100 K dust","Broader selection finds more little red dots and warm dust"],"cache_read_input_tokens":40832,"weakest_assumption_plain":"The selection box is drawn from color-redshift tracks of five LRD models; if real LRDs at 3<z<9 have colors outside that box, the ×1.7 number-density gain is a calibration artifact, not a true census increase.","fun_headline_variants_meta":{"raw":{"variants":["Little red dot census jumps 70% with new color selection","First warm dust emission seen in a little red dot","New photometric cut uncovers 1.7× more little red dots","JWST survey: 248 little red dots, one with 50–100 K dust","Broader selection finds more little red dots and warm dust"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000749,"raw_usage":{"total_tokens":3506,"prompt_tokens":1283,"completion_tokens":2223,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":899,"completion_tokens_details":{"reasoning_tokens":2131}},"tokens_in":899,"tokens_out":2223,"duration_ms":17960,"temperature":1.0,"reasoning_tokens":2131,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-12T00:52:24.739247+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"A MIRI-selected, spectroscopically confirmed sample of LRDs at z≈3–5 with F200W−F444W<1 mag, or a demonstration that the five model tracks misplace the observed LRD locus in Figure 1, would falsify the completeness claim. Alternatively, a complete F770W-selected census in the same fields yielding a surface density consistent with earlier samples (~0.5 arcmin⁻²) would show the ×1.7 gain is a selection artifact.","supporting_citations":[],"review_version":1}