{"id":"701acf1d-5ccb-440d-a310-2881410bbced","arxiv_id":"2504.17208","paper_version":2,"verdict":"REJECT","confidence":"MODERATE","novelty_score":4.0,"correctness_risk":"high","formal_verification":"none","parameter_count":3,"one_line_summary":"A Kolmogorov randomness analysis of JWST galaxy peak wavelengths finds a transition in their statistical distribution near redshift 2.7.","lead":"Using a statistical randomness test on the brightest spectral line of 148 JWST galaxies, the authors report a change in that line's distribution at redshift 2.7. The finding is a single-feature statistical signal, and the physical cause is not identified.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The claimed >99% confidence break at z≈2.7 rests on pointwise mock comparisons over heavily overlapping windows with a post hoc break choice, and the paper's stated assumption of redshift-independent noise is untested.","rationale":"After reading the manuscript, I agree with rejection but identify the statistical calibration rather than the noise assumption as the primary weak point. The pipeline produces a highly correlated KSP curve, and the claimed 99% confidence is based on pointwise mock intervals and a post hoc break redshift; no test accounts for the look-elsewhere effect. The permutation test described above would settle this concern. The self-stated assumption of identical noise and systematics (Section 5) is also untested and is a genuine risk, especially because the rest-frame 656 nm peak shifts across the NIRSpec band with redshift; however, a properly calibrated change-point test is logically prior, because without it the claim of a 99% detection fails even if the systematics were perfectly controlled. The use of mock-data comparison is a reasonable step, but it is insufficient for the global claim. The current evidence supports at best a tentative hint, not the stated >99% confidence detection, so the rejection verdict stands.","tokens_in":4469,"tokens_out":8148,"duration_ms":85778,"concrete_test":"Run a permutation-based scan test: randomly permute the galaxy redshifts among the 148 galaxies, keeping the measured rest-frame peak wavelengths fixed, and rerun the full pipeline (1000 nonidentical redshift intervals, KSP calculation with the generalized-normal null and normalization, and the Δz=0.1 moving average) for several thousand permutations. For each permutation, record the maximum absolute deviation of the smoothed real-data KSP from the mock median over the full redshift range. Compare the observed maximum near z≈2.7 to this empirical distribution; if fewer than 1% of permutations produce a maximum at least as large, the 99% claim survives the multiple-testing objection, and otherwise it does not.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The paper's headline claim is that the KSP of rest-frame peak wavelengths changes at z≈2.7 at >99% confidence. The confidence is derived by comparing the observed KSP sequence, built from 1000 heavily overlapping redshift intervals each containing only 10–20 of the 148 galaxies, with mock-data medians and 99% pointwise intervals, then smoothing with a Δz=0.1 moving average (Figs. 2–4). This procedure does not support the stated confidence: the KSP values are strongly correlated across overlapping intervals, the pointwise intervals ignore that correlation, and the break redshift is selected post hoc from the smoothed curve without a change-point scan or look-elsewhere correction. Under the null, a 4σ excursion somewhere in a long correlated sequence is not a 4σ detection. A second, independent problem is the paper's own stated requirement (Section 5) that 'instrumental noise and certain systematics are identical for the galaxies of the dataset'; this is not tested. Because the observed-frame position of the rest-frame 656 nm peak moves from ~1.9 μm at z=1.86 to ~5.3 μm at z=7.05, detector sensitivity, resolution, and calibration can vary across the sample and shift the estimated peak-wavelength distribution, producing a spurious break. Either issue alone is sufficient to undermine the central claim.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"This Letter applies the Kolmogorov stochasticity parameter (KSP) to JWST/NIRSpec spectra of 148 galaxies from the UNCOVER survey, selected to have their spectral maximum near rest-frame 656 nm and spanning redshifts 1.86–7.05. For 1000 random redshift intervals each containing 10–20 galaxies, the authors compute the KSP of the normalized wavelength values and compare it with mock data drawn from a generalized normal distribution with unit variance. They report deviations up to 4σ, smooth the KSP sequence with a moving average of Δz = 0.1, and conclude that the spectral properties change at z ≃ 2.7 at over 99% confidence level, which they attribute to possible evolution of galaxies or the intergalactic medium.","tokens_in":4690,"tokens_out":5176,"duration_ms":47787,"significance":"If the claimed break at z ≃ 2.7 were robust, it would point to a genuine redshift-dependent change in the distribution of rest-frame spectral peaks in high-redshift JWST galaxies, with potential implications for galaxy evolution or IGM studies. The paper uses a publicly available dataset and a well-defined statistical quantity, the KSP, and it makes a concrete, falsifiable prediction about where the change occurs. However, the central claim is not supported by the statistical analysis as presented: the significance estimate ignores the strong correlation among overlapping redshift windows, the break redshift is selected post hoc, the null distribution is under-specified, and the crucial assumption of redshift-independent instrumental noise is stated but untested. As a result, the paper currently does not provide convincing evidence for the claimed effect.","major_comments":[{"comment":"The shape parameter of the generalized normal distribution is never specified. The text states that a 'generalized normal distribution with variance equal to 1' is used as the theoretical distribution and that there is 'a single free parameter representing the sharpness of the distribution,' but the value of that parameter is not given. Because the mock-data comparison in Figs. 2–4 is the only basis for the reported 99% confidence, the entire null distribution is under-defined and the analysis cannot be reproduced.","section":"Section 4"},{"comment":"The 1000 redshift intervals are heavily overlapping, with each containing only 10–20 of the 148 galaxies. The KSP values in adjacent intervals are therefore strongly correlated, yet the 99% confidence intervals and the σλ used to compute Δλ/σλ in Fig. 3 are pointwise. A 4σ excursion somewhere in a long, correlated sequence is not a 4σ detection; the paper provides no correction for the effective number of independent intervals or for multiple testing.","section":"Section 4, Figs. 2–3"},{"comment":"The transition redshift z ≃ 2.7 is identified only after inspecting the smoothed KSP curve. No change-point test is performed, and no look-elsewhere penalty is applied. The paper's headline statement that the change is at 'over a 99% confidence level' is therefore not supported: the redshift is effectively a free parameter chosen to maximize the deviation, so the quoted confidence is circular.","section":"Section 4, Fig. 4 and Section 5"},{"comment":"The conclusion explicitly requires that 'the instrumental noise and certain systematics are identical for the galaxies of the dataset,' but this assumption is not tested. Because the rest-frame 656 nm peak shifts from about 1.9 μm at z = 1.86 to about 5.3 μm at z = 7.05, the observed-frame position moves across substantially different JWST/NIRSpec sensitivity, resolution, and calibration regimes. Redshift-dependent systematics could easily produce a spurious apparent break, and the analysis offers no control, such as splitting the sample by wavelength or comparing against galaxies with different rest-frame lines.","section":"Section 5"},{"comment":"The normalization of wavelength values to zero mean and unit variance before comparison with a fixed theoretical distribution invalidates the use of the Kolmogorov distribution Φ(λ) in Eq. (2) as the null distribution, because the data are no longer i.i.d. samples from the assumed distribution with known parameters. The paper does not state whether the mock data are normalized in the same way; if they are not, the null is mismatched, and if they are, the effective null distribution differs from the analytical KSP distribution due to the estimation of mean and variance. Either way, the quoted significance levels are not justified.","section":"Section 4"}],"minor_comments":[{"comment":"The text contains a typo: 'di fference' should be 'difference'.","section":"Section 4, sentence after Eq. (3)"},{"comment":"The reference 'Roberston et al 2023' appears in the text, but the reference list contains 'Robertson B.E., Tacchella S., Johnson B.D. et al, 2022'; the spelling and year should be corrected.","section":"Introduction"},{"comment":"The caption reads 'horizontal error bars indicate the selected [z1; z2] intervals,' but these are the horizontal segments of the blue points; the term 'error bars' is misleading. Clarify the visualization.","section":"Fig. 2 caption"},{"comment":"The phrase 'at over a 99% confidence level' is not formally defined as a frequentist confidence interval; specify whether this is a pointwise significance level after smoothing, and state the exact test used.","section":"Abstract and Section 5"}],"recommendation":"reject","confidential_remarks":"The statistical issues here are not merely cosmetic. The manuscript's central claim rests on a post hoc break identification combined with heavily overlapping pointwise comparisons, and the stated assumption of homogeneous instrumental systematics is both untested and implausible given the redshift-dependent wavelength shift. Even if the shape parameter were supplied and the multiple-testing issue were repaired, the lack of any systematics test would remain a fundamental obstacle. The paper would need a substantially different analysis, not just local revisions, to establish the claimed effect."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Colleague,\n\nThe paper applies the Kolmogorov stochasticity parameter to 148 JWST galaxies and claims a >99% confidence change in the distribution of rest-frame peak wavelengths at z≈2.7. The empirical claim is new—no one has run KSP on JWST spectra before—but the statistical support does not hold up.\n\nWhat it does well: it is a straightforward, clearly written application of a known technique. The authors are honest that the KSP only makes sense if instrumental noise and systematics are identical across the sample; they just don't test that condition. Using the 656 nm rest-frame peak as a tracer is a sensible choice, and the data selection (secure/solid redshift flags) is reasonable.\n\nThe soft spots are serious. First, the 1000 redshift intervals overlap heavily and each contains only 10–20 galaxies. KSP values across overlapping windows are strongly correlated, so the 99% pointwise intervals on the mock medians are not a valid basis for a global claim. A 4σ excursion somewhere in a long correlated sequence is not a 4σ detection. The break redshift is chosen after smoothing, with no change-point scan or look-elsewhere correction. Second, the paper's own load-bearing assumption—identical noise and systematics—is almost certainly false: the rest-frame 656 nm line moves from 1.9 μm at z=1.86 to 5.3 μm at z=7.05, crossing NIRSpec's sensitivity and calibration variations. That alone could produce a spurious transition at some redshift. Third, the generalized normal null distribution is said to have one free shape parameter, but the paper never says how it is estimated or whether the data constrain it.\n\nIndividually these are fixable; together they mean the central claim, as stated, is not supported. I'd still send this to peer review—the dataset is real, the method is legitimate, and a referee could ask for the missing controls. But I would not accept the paper until the analysis accounts for correlated windows and tests the noise assumption. As it stands, this is a tentative hint worth a follow-up, not an established result.","headline":"A clean application of KSP to JWST spectra, but the claimed z≈2.7 break is unsupported by the statistics as presented.","tokens_in":5265,"tokens_out":2216,"would_cite":false,"duration_ms":20553,"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":"Using the Kolmogorov stochasticity parameter on 148 JWST galaxy spectra, this paper claims the randomness properties of the peak-wavelength distribution change at z≈2.7 with >99% confidence.","keywords":["Kolmogorov stochasticity parameter","galaxy spectra","JWST NIRSpec","redshift evolution","UNCOVER survey","spectral peak wavelengths","generalized normal distribution","intergalactic medium"],"falsifier":"Re-run the Kolmogorov analysis on mock spectra that place the same underlying emission line at each observed wavelength with realistic NIRSpec noise; if the z≈2.7 excursion also appears when the input signal has no redshift dependence, the claimed transition is an artifact of noise or calibration.","tokens_in":4229,"feed_emoji":"🔭","tokens_out":6487,"duration_ms":58120,"temperature":0.7,"pith_summary":"This paper uses the Kolmogorov stochasticity parameter, a statistic that measures how far a sorted sample's cumulative distribution departs from a reference distribution, to search for redshift-dependent changes in JWST galaxy spectra. Applying it to the rest-frame wavelengths of the brightest spectral peaks of 148 galaxies from the UNCOVER NIRSpec survey, the authors claim that the randomness properties of this wavelength distribution change sharply at redshift about 2.7, at over 99 percent confidence. If the claim holds, it indicates that the balance between regular and stochastic components of these galaxy spectra is not constant across cosmic time, and that something—either in the galaxies or in the intergalactic medium—alters the signal near that redshift. The authors treat the result as a new signature to be confirmed with other data and methods.","feed_headline":"Spectral randomness shifts at z≈2.7 in JWST galaxies","feed_subtitle":"A 148-galaxy sample shows the peak-wavelength distribution leaves its baseline near z≈2.7 at >99% confidence.","key_machinery":"The machinery is the Kolmogorov stochasticity parameter, defined as λ_n = sqrt(n) sup_x |F_n(x) - F(x)|, where F_n is the empirical cumulative distribution of the sorted sample and F is a theoretical cumulative distribution. Kolmogorov's theorem gives a universal limiting distribution for this statistic that is independent of F, allowing it to serve as an objective measure of comparative randomness. The paper applies it to the normalized wavelength sequence in redshift windows, generates one thousand mock samples from a generalized normal distribution to define the expected value and the 99 percent confidence interval, and uses a moving average with Δz = 0.1 to locate the transition.","core_discovery":"The central claim is that the Kolmogorov stochasticity parameter of the rest-frame peak wavelengths, normalized per redshift interval to zero mean and unit variance and compared against a generalized normal distribution, departs from the mock-data expectation in a redshift-dependent way. The departure grows to about 4 sigma and, after moving-average smoothing over a redshift window of 0.1, marks a transition at z≈2.7. The authors interpret this as a change in the relative weight of random and regular sub-signals in the galactic spectra at that redshift, possibly tied to the intergalactic medium or to galaxy evolution. They report the result at over 99 percent confidence.","pith_inferences":["A direct test of the interpretation is to compute the same statistic on other rest-frame lines, such as [O III] or H-beta, in the same galaxies; if the z≈2.7 transition reflects changing interstellar-medium conditions, it should also appear in those lines, while a wavelength-dependent instrumental effect would shift or vanish.","The 148-galaxy sample spans a wide range of signal-to-noise; removing the lowest-quality spectra or requiring secure redshift flags would show whether a few faint objects drive the 4-sigma excursion.","Because the Kolmogorov statistic is computed from peak wavelengths only, a redshift-dependent mix of galaxy types, such as active galactic nuclei versus star-forming galaxies, could mimic the signal; classifying galaxies independently would distinguish population mixing from a true spectral change."],"forward_implications":["The spectral peak wavelength distribution of JWST galaxies is not statistically stationary across the redshift range 1.86 to 7.05.","A change at z≈2.7 means the random-regular mix in the emission lines differs at higher redshift, so galaxy samples above and below that redshift should not be pooled without accounting for the difference.","If astrophysical in origin, the transition provides a new redshift marker for galaxy or intergalactic-medium evolution that is independent of traditional photometric or color diagnostics.","The same Kolmogorov analysis can be applied to other emission lines and to other deep surveys to test whether the transition is line-specific or a general spectral property."],"supporting_citations":[{"why":"Supplies the theorem that the limiting distribution of the Kolmogorov stochasticity parameter is universal and independent of the underlying cumulative distribution.","marker":"Kolmogorov 1933"},{"why":"Establishes the stochasticity parameter as a tool for comparing the randomness properties of cumulative signals and documents its behavior for small data sequences.","marker":"Arnold 2008a,b"},{"why":"Provides the UNCOVER galaxy sample, the redshift-quality flags ('secure' and 'solid'), and the 148 galaxies with rest-frame spectral maxima near 656 nm.","marker":"Price et al 2024"},{"why":"Describes the UNCOVER survey from which the spectroscopic data are drawn.","marker":"Bezanson et al. 2024"},{"why":"Is the reference for the JWST NIRSpec instrument that produced the galaxy spectra analyzed here.","marker":"Jakobsen et al. 2022"},{"why":"Provides the NIRSpec instrument description and calibration context used by the authors for interpreting the measured spectra.","marker":"McElwain et al. 2023"}],"fun_headline_variants":["JWST galaxies reveal spectral shift at z≈2.7","Kolmogorov analysis spots z≈2.7 spectral break","Galaxy spectra change at z≈2.7 in JWST survey","JWST data: spectral properties transition at z≈2.7","Spectral break at z≈2.7 in JWST galaxies"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The result depends on detector noise and calibration systematics being identical across the entire redshift range, so that the only thing changing with redshift is the galaxy signal itself.","fun_headline_variants_meta":{"raw":{"variants":["JWST galaxies reveal spectral shift at z≈2.7","Kolmogorov analysis spots z≈2.7 spectral break","Galaxy spectra change at z≈2.7 in JWST survey","JWST data: spectral properties transition at z≈2.7","Spectral break at z≈2.7 in JWST galaxies"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000273,"raw_usage":{"total_tokens":1562,"prompt_tokens":798,"completion_tokens":764,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":414,"completion_tokens_details":{"reasoning_tokens":673}},"tokens_in":414,"tokens_out":764,"duration_ms":7027,"temperature":1.0,"reasoning_tokens":673,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-16T10:46:11.729101+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Re-run the Kolmogorov analysis on mock spectra that place the same underlying emission line at each observed wavelength with realistic NIRSpec noise; if the z≈2.7 excursion also appears when the input signal has no redshift dependence, the claimed transition is an artifact of noise or calibration.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the theorem that the limiting distribution of the Kolmogorov stochasticity parameter is universal and independent of the underlying cumulative distribution."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Describes the UNCOVER survey from which the spectroscopic data are drawn."},{"cited_title":"et al, 2022, A&A, 661, A80","cited_arxiv_id":null,"evidence_quote":"Is the reference for the JWST NIRSpec instrument that produced the galaxy spectra analyzed here."},{"cited_title":"W., Feinberg L.D., Perrin M.D","cited_arxiv_id":null,"evidence_quote":"Provides the NIRSpec instrument description and calibration context used by the authors for interpreting the measured spectra."}],"review_version":1}