{"id":"b0a9f415-aa3e-4529-be24-4938f705fce6","arxiv_id":"2501.08568","paper_version":3,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":5,"one_line_summary":"About 67% of cosmic noon LAEs are first-burst galaxies, 28% have dominant current bursts with past star formation, and 5% had their biggest burst in the past; overall, 95% are having their largest burst yet.","lead":"This paper reconstructs the star formation histories of 74 Lyman alpha emitting galaxies at cosmic noon and finds that about 95% are currently experiencing their largest star formation burst, while a third show significant past star formation. It challenges the simple picture of LAEs as newborn galaxies and implies that multiple evolutionary paths can create strong Lyman alpha emission.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Untested prior sensitivity in Dense Basis reconstruction is the main risk: the Dirichlet α=3 and logSFR bounds can shift past SFRs across the 1 Msun/yr threshold, changing the 67/28/5 archetype fractions.","rationale":"Reader's weakest assumption is on the mark. The paper has real strengths: a previously published non-parametric SFH method, a large narrowband-selected sample, and an internal check that the χ² cut moves the archetype fractions by only ~10% while preserving the 95% formative fraction (§4.2). Those make the result credible at the level of 'most LAEs are in a formative burst.' My concern does not overturn that. It targets the more specific 67/28/5 taxonomy and the interpretation that a substantial minority of LAEs had significant past star formation. That interpretation depends on recovering whether past SFR crossed ~1 Msun/yr — a low bar close to the noise floor of SED-based SFH reconstruction. The Dense Basis validation in Iyer et al. (2019) is encouraging, but it was not demonstrated for this exact redshift/filter/depth combination, and the paper's only prior check (varying the number of time bins) does not explore the shape parameter α or the SFR range. A mock-recovery prior-sensitivity test is the natural, decisive check: it directly measures how often reconstructed SFHs cross the threshold when the truth is known. If the fractions are stable under this test, the conditional verdict can be upgraded; if not, the paper should present the fractions with prior-averaged uncertainties or soften the taxonomy. This is exactly the conditional-verdict situation the reader identified, so I recommend no change to the verdict.","tokens_in":18063,"tokens_out":5486,"duration_ms":57026,"concrete_test":"Run a mock-recovery test using the exact ODIN/UVCANDELS filter set and photometric depth: generate ~200 LAEs at z=2.4, 3.1, 4.5 from known Dense Basis SFHs spanning all three archetypes, with past SFRs deliberately straddling 1 Msun/yr (0.3, 0.7, 1.0, 1.5, 3 Msun/yr). Add realistic noise at UVCANDELS depths and re-fit with Dense Basis using the fiducial priors and variants: Dirichlet α = 1, 2, 3, 5 and logSFR priors [-2,3] and [-1,2] (and, if feasible, an independent code such as Prospector). Report the confusion matrix and the recovered first/dominant/non-dominant fractions for each variant. Stability of the fractions within a few percentage points would retire the concern; shifts above 10 points would show the archetype fractions are prior-dominated and the central claim needs re-scoping.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Central claim: 67% first-burst, 28% dominant-burst, 5% non-dominant-burst, hence ~95% formative-burst LAEs. The first/dominant distinction is defined by whether the reconstructed SFR exceeded ~1 Msun/yr at any past epoch (§4.1). This boundary sits exactly where SED fitting is weakest: rest-UV-through-NIR photometry constrains the current SFR and total stellar mass well, but a past burst of 1 Msun/yr contributes only a modest old stellar population and can be absorbed into the continuum or suppressed by the prior. The adopted priors in §3.1 include a Dirichlet prior with α=3 on the mass fractions in time bins and a uniform logSFR prior [-1,2]. α=3 is a smoothing prior relative to α=1; it can suppress short, low-amplitude past bursts. The paper reports that changing the number of time bins does not change classifications, but it does not vary α or the SFR prior bounds. Because the 1 Msun/yr threshold is set by the sample's own minimum time-of-observation SFR rather than an external calibration, even a 0.2–0.3 dex shift in reconstructed past SFRs can move objects between first and dominant. If past peaks are exaggerated above the current SFR, the non-dominant fraction and the 95% formative fraction would also change. Thus the headline fractions and the 'multiple evolutionary paths' conclusion are not yet protected against prior sensitivity.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The manuscript uses the Dense Basis non-parametric SED-fitting method to reconstruct star formation histories for 74 ODIN narrowband-selected LAEs at z=2.4, 3.1, and 4.5 with UVCANDELS rest-UV-through-NIR photometry. The authors define three SFH archetypes: 'first burst' (no significant past SFR), 'dominant burst' (significant past SFR but current SFR largest), and 'non-dominant burst' (largest SFR in the past). They report 67% first-burst, 28% dominant-burst, and 5% non-dominant-burst LAEs, combine the first two into a 'formative burst' label (~95%), and compare the fraction of stellar mass formed in the last 200 Myr with mass-matched LBG control samples, finding LAEs assemble ~1.3 times more of their mass recently. The central claim is that multiple stellar mass assembly histories can produce strong Ly-alpha emission, and that the simple young-first-burst picture of LAEs is incomplete.","tokens_in":18353,"tokens_out":6118,"duration_ms":63166,"significance":"If the archetype fractions are robust, this is a valuable observational result that moves beyond parametric SFH assumptions and quantifies the diversity of LAE mass assembly at cosmic noon. The comparison with LBG controls, including KS tests, is a concrete and falsifiable check on the claim that LAEs differ from other star-forming galaxies. The paper also contains useful self-checks, such as examining the magnitude and EW dependence of the chi-squared cut and explicitly reporting that including chi-squared-rejected objects shifts the archetype distribution by about 10 percentage points while preserving the ~95% formative-burst fraction. These strengths make the paper a serious candidate for publication, but the missing prior-sensitivity analysis and lack of propagated uncertainties on the archetype fractions currently leave the headline percentages insufficiently protected.","major_comments":[{"comment":"The boundary between 'first' and 'dominant' bursts is defined by a threshold of SFR ≥ 1 M☉/yr (§4.1), which is set equal to the sample's minimum time-of-observation SFR. The classification is based on the median reconstructed SFH, yet the paper does not test how the 67/28/5 archetype fractions respond to the adopted Dense Basis priors. The only sensitivity test reported in §3.1 varies the number of t_x time bins; the Dirichlet concentration α=3.0 and the logSFR prior range [−1, 2] are held fixed. Because the reconstruction scatter quoted from Iyer et al. (2019) is 0.2 dex, individual objects can plausibly cross the 1 M☉/yr boundary if α or the SFR bounds change. I request a sensitivity analysis varying α (e.g., 1, 2, 4) and the SFR prior bounds, with the resulting first/dominant/non-dominant fractions and the combined ~95% formative-burst fraction reported for each setup.","section":"Section 4.1 (with §3.1)"},{"comment":"The archetype percentages (77%, 57%, 59%, etc.) are quoted as point values with no propagated uncertainties, even though each object's SFH is a posterior distribution and the 25–75% intervals are shown in Figures 1–4. Since the three archetypes are mutually exclusive, Poisson binomial counting uncertainties are straightforward to compute. A more complete treatment would classify each object by the fraction of posterior SFH samples satisfying each archetype condition (e.g., max(past SFR) ≥ 1 M☉/yr and max(past SFR) < current SFR) and propagate those per-object probabilities into the reported fractions. Without this, the reader cannot tell whether the 28% dominant-burst fraction is a robust population signal or a threshold artifact of the median SFH.","section":"Section 4.1"}],"minor_comments":[{"comment":"The 'formative burst' fraction is written as '~95' and '∼95' without the percent sign; it should read '~95%' for clarity.","section":"Abstract and Section 5 (conclusion 4)"},{"comment":"In the sentence describing priors, 'the log of the instantaneous star formation rateSF R' has a missing space and the formatting of 'SF R' should be corrected to 'SFR'.","section":"Section 3.1"},{"comment":"The notation 'M ed(M∗‘NOW’/M∗)' is awkward and contains typographic quotes; define a clean symbol such as M∗,now/M∗ and use it consistently throughout the text and figures.","section":"Section 4.2"},{"comment":"The choice of a 200 Myr resolution for identifying the SFR peak is attributed to the near-UV SFR timescale of Broussard et al. (2019), but the connection between that timescale and the peak-search bin width could be stated more explicitly.","section":"Section 3.3"}],"recommendation":"major_revision","confidential_remarks":"The stress-test concern about prior sensitivity is legitimate and lands on a load-bearing point. The paper itself already contains a useful self-check on the chi-squared cut (Section 4.2), which mitigates one of the skeptic's worries, but the lack of variation of α and the SFR prior bounds, plus the absence of posterior-probability-based classification, are the key technical gaps. The manuscript is well within the scope of the journal and the central idea is publishable after a focused revision. I would not require a full new observational dataset; a prior-sensitivity appendix and per-object classification probabilities would suffice."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The genuinely new thing here is the first non-parametric star formation history census of LAEs at cosmic noon. Previous work used parametric forms and couldn't separate first-burst from rejuvenated systems. This paper shows that while 67% of 74 LAEs are in their first major burst, 28% have significant past star formation with the current burst dominant, and 5% are non-dominant — so the old picture of LAEs as uniformly first-burst galaxies is incomplete. That is a real result, and the sample construction is careful: rest-UV-through-NIR photometry, spec-z validation, a chi-squared cut tested for EW and magnitude bias, and the Dense Basis method with published 0.2 dex reconstruction accuracy out to 5 Gyr. The formative-burst category (first + dominant, ~95%) is the most robust claim; even the paper's own chi-squared-cut test shifts the split by ~10 points while keeping ~95% formative.\n\nThe soft spots are real but addressable. Most importantly, the first-versus-dominant boundary rests on whether the reconstructed SFR ever exceeded ~1 Msun/yr, and that threshold is set by the sample's minimum time-of-observation SFR. The paper tests the number of time bins but does not vary the Dirichlet alpha or the logSFR prior bounds. Since 1 Msun/yr sits right where SED fitting is weakest, the 67/28/5 split should be treated as indicative, not precise. Second, the archetype fractions have no propagated uncertainties. With 74 galaxies the Poisson errors alone are several percentage points per bin; per-redshift fractions like 77%/57%/59% have even larger error bars. Third, the abstract overstates the LBG comparison: at z=2.4 the KS test gives p=0.80, so “1.3 times higher” is hiding a null at the lowest redshift, and the claim that most LBGs are dominant-burst is false for the z=2.4 control sample. These are minor-to-moderate and fixable with a sensitivity analysis, bootstrap uncertainties, and a more careful abstract.\n\nThe non-dominant burst subsample is just 4 objects, and the authors appropriately flag them as case studies. The Dense Basis citations are legitimate; Iyer et al. 2019 is an external published method, and self-citation here is not a red flag.\n\nThis paper deserves a serious referee. It is a solid, honest observational study with a new measurement that will be cited. My recommendation: send it to peer review, and ask for a prior sensitivity test, uncertainty propagation on the archetype fractions, and an abstract that matches the z=2.4 null.","headline":"A careful first non-parametric SFH census of cosmic-noon LAEs, with a useful taxonomy whose exact fractions are softer than the headline implies.","tokens_in":19062,"tokens_out":3064,"would_cite":true,"duration_ms":30626,"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":"Most Lyman-alpha galaxies are experiencing their largest starburst yet, but a third have star-forming pasts.","keywords":["Lyman alpha emitters","star formation histories","Dense Basis","non-parametric SED fitting","cosmic noon","galaxy evolution","starbursts","ODIN survey"],"falsifier":"Re-run the same 74 LAEs through an independent non-parametric SFH code or vary the Dirichlet prior parameter alpha and the SFR prior, then recompute the first/dominant/non-dominant fractions; if the ~95% formative-burst result shifts by more than the quoted statistical uncertainties, the classification is a prior artifact rather than a property of the galaxies.","tokens_in":17845,"feed_emoji":"🌌","tokens_out":6942,"duration_ms":60708,"temperature":0.7,"pith_summary":"This paper tests a long-standing assumption about Lyman Alpha Emitting galaxies (LAEs): that they are observed during their very first major burst of star formation. Using the Dense Basis method to reconstruct non-parametric star formation histories from rest-UV-through-NIR photometry of 74 ODIN LAEs at redshifts 2.4, 3.1, and 4.5, the authors find that 67% of LAEs match the first-burst archetype, 28% have had significant past bursts with the current burst still largest, and 5% had their largest burst in the past. Combining the first two classes, roughly 95% of LAEs are experiencing their largest burst yet, which they call a formative burst. The authors also find that LAEs built a larger fraction of their stellar mass in the last 200 Myr than mass-matched Lyman Break Galaxies. If correct, strong Ly-alpha emission does not require a unique, first-time starburst, and multiple evolutionary paths can produce LAEs.","feed_headline":"About 95% of Lyman-alpha galaxies are in their biggest burst yet","feed_subtitle":"Reconstructed histories show a third of LAEs had past starbursts, so multiple paths can make a strong Ly-alpha emitter.","key_machinery":"The load-bearing tool is Dense Basis, a non-parametric, Gaussian-process-based method for reconstructing star formation histories from spectral energy distributions. Unlike parametric SFH models that assume a single functional form and tend to place the only peak at the time of observation, Dense Basis fits flexible histories with multiple possible bursts, allowing the authors to identify where the maximum star formation rate falls relative to the observation epoch and to separate first, dominant, and non-dominant burst classes. The classification itself uses a 200 Myr resolution at the time of observation and a 1 solar mass per year threshold for what counts as a significant past burst.","core_discovery":"The paper's central claim is that LAE star formation histories come in three archetypes, not one. From Dense Basis reconstructions, the authors classify 67% of their 74 LAEs as first-burst galaxies, with only modest star formation before observation; 28% as dominant-burst galaxies, where an earlier significant burst occurred but the current burst has the highest star formation rate; and 5% as non-dominant-burst galaxies, where the peak star formation rate lies in the past. Because the first and dominant classes together make up about 95% of the sample, the authors argue that almost all LAEs are undergoing a formative burst, a burst that plays a major role in building the galaxy. They further report that LAEs have a median fraction of stellar mass formed in the last 200 Myr about 1.3 times higher than mass-matched LBGs, and that most LBG controls show dominant-burst histories. The conclusion is that being a strong Ly-alpha emitter is not evidence of being a first-time starburst, and galaxies can become LAEs through several mass-assembly pathways.","pith_inferences":["Beyond the paper, if LAEs can be in dominant bursts, then LAE-based estimates of star formation rate density or clustering may need to account for a fraction of galaxies that are not pristine first-time starbursts.","The non-dominant LAEs, though rare, offer a direct test of whether Ly-alpha escape depends more on dust and ISM geometry than on current star formation rate.","A testable extension would be to compare Dense Basis classifications with rest-frame optical line ratios or resolved stellar population ages for the same galaxies to see if past bursts leave independent signatures.","The paper's definition of a 'significant' past burst uses a fixed 1 solar mass per year threshold tied to the sample's own minimum SFR; an external threshold from Ly-alpha luminosity or stellar mass could change the archetype fractions."],"forward_implications":["A significant minority of LAEs (about one-third) have had substantial past star formation, so first-burst is not a necessary condition for strong Ly-alpha emission.","About 95% of LAEs are experiencing their largest burst of star formation to date, meaning a formative burst is the norm even when the history is not a first burst.","LAEs assemble a larger fraction of their stellar mass in the most recent 200 Myr than mass-matched LBGs, confirming that LAEs have less evolved stellar populations.","A majority of LBG controls show dominant-burst rather than first-burst histories, pointing to a systematic difference in mass assembly between LAEs and other star-forming galaxies.","Because roughly a third of LAEs once had SFRs above the current sample minimum, LAE status may be temporary and repeatable within a galaxy's lifetime."],"supporting_citations":[{"why":"Introduces the Dense Basis method for flexible, non-parametric star formation history reconstruction.","marker":"Iyer & Gawiser 2017"},{"why":"Validates Dense Basis on simulated galaxies and establishes its accuracy in recovering multiple SFH peaks.","marker":"Iyer et al. 2019"},{"why":"Defines the ODIN LAE selection, Ly-alpha equivalent-width measurement, and interloper rejection that produce the sample.","marker":"Firestone et al. 2024"},{"why":"Establishes the ODIN survey design and narrowband filters that set the LAE redshifts and redshift priors.","marker":"Lee et al. 2024"},{"why":"Provides the UVCANDELS photometric catalog and UV data used in the SED fits.","marker":"Wang et al. 2025"},{"why":"Supplies the photometric redshifts used to construct the mass-matched LBG control samples.","marker":"Mehta et al. 2024"},{"why":"Introduces the Dirichlet prior and non-parametric SFH parameterization adopted in the fiducial Dense Basis runs.","marker":"Leja et al. 2017"},{"why":"Documents the bias of parametric SFH models and informs the priors used in the non-parametric setup.","marker":"Leja et al. 2019"}],"fun_headline_variants":["95% of Lyman-alpha galaxies are in their largest burst","One-third of Lyman-alpha galaxies had earlier starbursts","LAE archetypes: first, dominant, or past-peak bursts","Three star-formation paths to Lyman-alpha emission","Not all LAEs are first-timers: 33% have prior bursts"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The classification assumes that the Dense Basis reconstruction, with its chosen priors, can reliably tell whether a galaxy's past star formation rate ever rose above about one solar mass per year.","fun_headline_variants_meta":{"raw":{"variants":["95% of Lyman-alpha galaxies are in their largest burst","One-third of Lyman-alpha galaxies had earlier starbursts","LAE archetypes: first, dominant, or past-peak bursts","Three star-formation paths to Lyman-alpha emission","Not all LAEs are first-timers: 33% have prior bursts"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000907,"raw_usage":{"total_tokens":3968,"prompt_tokens":1084,"completion_tokens":2884,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":700,"completion_tokens_details":{"reasoning_tokens":2793}},"tokens_in":700,"tokens_out":2884,"duration_ms":20894,"temperature":1.0,"reasoning_tokens":2793,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-10T20:23:13.543220+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Re-run the same 74 LAEs through an independent non-parametric SFH code or vary the Dirichlet prior parameter alpha and the SFR prior, then recompute the first/dominant/non-dominant fractions; if the ~95% formative-burst result shifts by more than the quoted statistical uncertainties, the classification is a prior artifact rather than a property of the galaxies.","supporting_citations":[],"review_version":1}