{"id":"7818ae70-a9ba-44f0-a0c5-827730bb12ff","arxiv_id":"2411.16023","paper_version":3,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":5.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":4,"one_line_summary":"The minimum-entropy pulse profile gives dispersion measures matching catalog values for 45 of 48 FAST pulsars, offering a simple model-free DM indicator.","lead":"This paper tests a new way to find the dispersion measure (DM) of pulsars and fast radio bursts: compute the Shannon entropy of the pulse profile for many trial DMs and pick the one with minimum entropy. On 48 pulsars observed with FAST, the resulting DMs mostly agree with published values, suggesting a simple, model-free alternative to template fitting.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The 3-sigma clipping and RFI-removal preprocessing in §2.1 can bias the entropy minimum, and with no sensitivity analysis the reported DM agreement is not yet shown to be a property of the entropy indicator itself.","rationale":"The reader's weakest assumption identified exactly this preprocessing dependency, and I agree with it. The paper's own §2.1 states that the 3σ retention and the three-segment interference removal are only for reference and are assumed to preserve the true signal, but no test of that assumption is provided. Since the Appendix A code applies the 3σ cut before dedispersion and then computes entropy from the surviving samples, the indicator is not a pure entropy of the observed profile; it is an entropy of a thresholded version. The theoretical claim in §1 that the min-entropy profile is 'closest to the real distribution' therefore goes beyond what is demonstrated. However, the paper's practical DM-recovery claim is supported by 48 real pulsars and does not require the optimality claim; the authors themselves disclaim optimality in the abstract and conclusions. A sensitivity analysis could either confirm or refute whether the preprocessing is responsible for the agreement. Until then, CONDITIONAL remains the appropriate verdict, and no change to the reader's recommendation is needed.","tokens_in":12276,"tokens_out":5239,"duration_ms":55985,"concrete_test":"Take a subset of the 48 FAST pulsars (or the synthetic J1857+0212 setup of §2.3) and recompute the min-entropy DM while varying the clipping threshold from 2σ to 4σ, with clipping disabled by subtracting a robust noise baseline before binning, and with the interference-removal step turned off. If any recomputed DM moves by more than its quoted error bar in Table 1, or the 0.5%-agreement count changes materially, the preprocessing assumption is load-bearing and the central claim requires qualification; if all DMs are stable, the concern is retired.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The paper's central empirical result (Table 1, §2.4) is obtained only after discarding every data point below 3σ in each slice and applying a three-segment interference-removal step that is assumed to leave the true pulse untouched (§2.1 and the Appendix A filter `allH['snr']>1+3*sig`). The entropy in `func()` is then computed on the surviving bright samples only. This is load-bearing because the entropy functional is evaluated on a truncated, pre-selected distribution rather than on the full flux profile. If a real pulse has low-amplitude wings, a scattering tail, or any component below the 3σ cutoff, those samples are silently removed, and the 'minimum entropy' DM is the DM that most concentrates the surviving bright core. That may still be a useful DM estimator, but it is not evidence for the stronger wording in §1 that the min-entropy profile is 'closest to the real distribution.' The reported agreement for 45 of 48 pulsars is encouraging, but since the threshold and the RFI-removal logic are fixed ad hoc and never varied, the result could be an artifact of these preprocessing choices rather than of the entropy criterion.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper proposes a template-free, entropy-based dispersion measure (DM) estimator: for a given trial DM, the dedispersed time series is binned into a flux profile, its Shannon entropy is computed, and the DM that minimizes this entropy is taken as the optimal DM. The authors argue that the minimum-entropy profile is closest to the true pulse profile and thus yields the correct DM. They apply the method to 48 pulsars observed with FAST, reporting that all but three entropy DMs differ from catalog reference DMs by less than 0.5%, and that the three outliers agree with older references. A synthetic-data test and an example code are included.","tokens_in":12473,"tokens_out":5669,"duration_ms":53313,"significance":"If the empirical agreement is robust, the paper offers a computationally cheap, template-free DM indicator that could complement existing fitting and profile-discrimination methods, particularly for studying single-pulse profiles and FRBs. The validation is against independent catalog DMs (not fitted to them), and the paper ships example code and a synthetic-data sanity check, which are genuine strengths. However, the central claim currently depends on several ad hoc preprocessing steps whose influence has not been quantified, so the significance is conditional on a sensitivity analysis.","major_comments":[{"comment":"The preprocessing is load-bearing: the entropy is computed only on data points surviving the 3-sigma cutoff (`allH['snr']>1+3*sig` in the code), and the three-segment interference-removal procedure is described only in prose, with the code leaving a placeholder ('add your function here'). Every entropy value is therefore evaluated on a truncated, threshold-selected subset of the flux distribution, not on the full pulse profile. If real pulses contain low-amplitude wings, scattering tails, or other components below the threshold, those samples are silently discarded and the minimum-entropy DM could shift. The paper provides no sensitivity analysis varying the threshold (e.g., 1, 2, 5 sigma, or no cut), the interference-removal logic, the phase offset, or the smoothing window. Without such a test, the agreement in Table 1 is not yet shown to be a property of the entropy indicator itself rather than of the preprocessing choices. Please add a sensitivity study, or at minimum report the fraction of data points retained by the 3-sigma cutoff per pulsar and demonstrate that the derived DM is stable under reasonable variations of these parameters.","section":"§2.1 and Appendix A"},{"comment":"The headline result is that 45 of 48 entropy DMs agree with reference DMs within 0.5%, but the comparison does not use the stated error bars. Several entropy DMs have uncertainties larger than the 0.5% band (e.g., J1404+1159 with err 2.8 on DM 15.857), and the reference DMs are quoted without their own uncertainties. In addition, the 'weighted average' of per-pulse DMs is not defined: the weights, the number of pulses used, and how outliers are combined are all unspecified. Please provide the per-pulse measurements, the weighting scheme, and the uncertainties, or clearly state that the 0.5% comparison is against the catalog central values only, so the reader can judge whether the differences are statistically significant.","section":"§2.4 and Table 1"},{"comment":"The conceptual premise that 'the flux distribution with the minimum entropy (maximum amount of information) is the closest to the real distribution' is asserted rather than derived, and the paper itself concedes in the Abstract and §3 that the criterion cannot be proven optimal. This is acceptable as a heuristic motivation, but the title and the introductory wording claim a direct relationship to the 'true profile.' Because the empirical validation in §2.4 is only against catalog DMs, not against independent measurements of the true pulse profile, I recommend rewording the claims to say that the minimum-entropy DM is a useful estimator that recovers reference DMs for the 48 FAST pulsars studied, and to present the 'closest to the real profile' statement as a motivation rather than a conclusion of the paper.","section":"§1 and Abstract"}],"minor_comments":[{"comment":"The heading 'Disscusions and Conclusions' contains a typo and should read 'Discussion and Conclusions.'","section":"Section 3 heading"},{"comment":"The caption says 'Using pulsar J1901+0331 as an example,' but the panel title and the surrounding text refer to J1922+1733 with DM=234.0; please correct this mismatch.","section":"Figure 2"},{"comment":"The error-estimation strategy does not specify the width of the smoothing window; the code uses binsize = (max(dms)-min(dms))/20, but this choice is not justified and can change the reported error range.","section":"§2.2"},{"comment":"For each pulsar, the 'References' column lists two references without indicating which one supplies the quoted reference DM; please split this into two columns or add a note explaining the provenance of each DM value.","section":"Table 1"},{"comment":"The '3 sigma' threshold is not defined precisely: it should be stated whether sigma is the standard deviation of the full slice, the median-normalized noise estimate used in the code, or another quantity.","section":"§2.1"},{"comment":"The sentence 'their are also some pulsars known as millisecond pulsars' contains a grammatical error ('their' should be 'there').","section":"§1"}],"recommendation":"major_revision","confidential_remarks":"The paper fits the scope of astro-ph.IM as a methods paper. The main risk is that the reported agreement is driven by the ad hoc preprocessing rather than by the entropy criterion; a sensitivity analysis is therefore essential before the claim can be accepted. No concerns about citation practices or novelty disclosure."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Colleague,\n\nThe paper proposes using Shannon entropy as a DM-selection criterion: dedisperse, bin, normalize, compute entropy, pick the DM that minimizes it. That specific indicator is new as far as I know, within the established profile-discrimination family (maximize fine structure, subband consistency, S/N). It is simple and template-free, and they test it on 48 FAST pulsars, getting within 0.5% of catalog DMs for all but three, and those three match older references. The authors are honest that they cannot prove the minimum-entropy profile is the true one; they recommend using multiple indicators. That honesty is a real strength.\n\nWhat is good: the method is easy to implement, the synthetic test checks the code for systematic errors, and the validation set is reasonably sized. The preprint also ships an example code, which helps reproducibility even though the data and full pipeline are not released. I don't see circularity: the entropy DMs are compared to independent catalog values, not fitted to them.\n\nThe soft spots are material but not fatal. First, there is no head-to-head comparison against existing indicators (S/N, DM_phase, subband consistency) on the same data. Without that, it's hard to know whether the min-entropy criterion adds anything or is just another way to pick a similar DM. Second, the preprocessing is load-bearing: data points below 3-sigma are discarded before entropy is calculated, and an RFI-removal step assumes the true pulse is never clipped. The stress-test note is right that this can bias the entropy minimum, and no sensitivity analysis is shown. I'd like to see the 3-sigma threshold varied, and the entropy recomputed on the full (unclipped) profile. Third, the error bars are heuristic, and the 'weighted average' of per-pulse DMs is not described in detail. These are fixable in revision.\n\nThe central empirical claim—that a cheap entropy scan recovers published DMs—is plausible, and the three outliers being consistent with older references is a good sign. This is not a breakthrough, but it is honest progress. I'd send it to a referee: it deserves a serious look, especially from someone who knows the profile-discrimination literature. The revision should add a baseline comparison and a sensitivity analysis; without those, the paper stays a useful note but not a definitive evaluation.\n\nRecommendation: engage with it, but ask for the missing comparisons and robustness checks.","headline":"Entropy-minimization as a DM indicator works reasonably on 48 FAST pulsars, but without a baseline or sensitivity run the agreement isn't yet shown to come from the entropy criterion.","tokens_in":13031,"tokens_out":2610,"would_cite":false,"duration_ms":23609,"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":"A minimum-Shannon-entropy scan over trial dispersion measures picks the profile closest to the true one; on 48 FAST pulsars it matches published DMs within 0.5% for all but three (which match older references).","keywords":["dispersion measure","Shannon entropy","pulse profile","pulsar","FAST telescope","profile discrimination","template-free","radio astronomy"],"falsifier":"Inject a synthetic dispersed pulse with a known DM whose true profile has a low-amplitude broad component below the 3-sigma cutoff: if the entropy-DM minimum moves away from the injected DM by more than the quoted error when that weak component is present, and does not move when the component is removed, the preprocessing assumption is falsified.","tokens_in":12012,"feed_emoji":"📡","tokens_out":9141,"duration_ms":79444,"temperature":0.7,"pith_summary":"The paper tries to establish a template-free rule for choosing the dispersion measure (DM) of a radio pulse: dedisperse at many trial DMs, bin each trial profile, and compute Shannon entropy; the trial DM with the lowest entropy is argued to give the profile closest to the real one. The authors test this on 48 pulsars observed with FAST, computing an entropy DM for each pulse and averaging per pulsar. All but three of their entropy DMs differ from the latest published values by less than 0.5%, and those three agree with older published DMs instead. If the rule holds, DM estimation no longer needs a pulse-shape model, which matters for studying real profile structure rather than forcing data into a template. The paper itself notes that the minimum-entropy criterion is not proven optimal and recommends comparing several indicators.","feed_headline":"Lowest-entropy scan recovers pulsar dispersion measures","feed_subtitle":"No templates or fitting: minimum entropy matches published DMs for 45 of 48 pulsars, with 3 matching older references.","key_machinery":"The load-bearing object is the entropy functional $H(X)=-\\sum_{i=1}^{n} p(x_i)\\log_2 p(x_i)$, where $x_i$ is a short time interval of the candidate pulse profile and $p(x_i)$ is the fraction of the total flux received in that interval. Minimizing this functional over trial dispersion measures is the whole algorithm: it replaces template fitting with a pure concentration score. Two auxiliary procedures carry the practical argument: the 3-$\\sigma$ retention of data points and the three-segment interference-removal routine that the paper assumes leaves true pulse signals intact, and the error strategy that smooths the entropy-DM curve and takes the 3-$\\sigma$ band around the minimum as the DM uncertainty.","core_discovery":"The central claim is that within a fixed time window, the flux distribution of a dedispersed pulse profile with minimum Shannon entropy is the distribution closest to the true profile, and the dispersion measure that produces it is the optimal DM. Concretely, for each trial DM the data are shifted by the frequency-dependent delay of Equation (1), binned in time, normalized to a probability distribution, and scored by $H=-\\sum_i p_i\\log_2 p_i$. The DM that minimizes $H$ is taken as the measured DM, with an error range read off from the entropy-versus-DM curve using a 3-$\\sigma$ smoothed-residual rule. The validation on 48 FAST pulsars compares these entropy DMs against the latest references: 45 are within 0.5%, and the 3 remaining pulsars match older references, which the paper interprets as consistency with published DM values once reference-systematic differences are allowed.","pith_inferences":["Editorial inference: the minimum-entropy score is essentially a sharpness or concentration measure, so it should behave much like structure-maximization dedispersion used in fast radio burst work; a direct comparison of min-entropy DM with structure-maximized DM on FRB data would be a natural test the paper does not run.","Editorial inference: since only 3-sigma-surviving samples enter the entropy, the indicator measures the bright core of the pulse, not its low-level wings; if scattering broadens a profile or a genuine component sits below threshold, the entropy minimum may drift, and a soft-threshold variant is the obvious check.","Editorial inference: the three outliers matching older references hint that published DMs carry their own processing systematics, so catalog agreement is a weak ground truth; a stronger validation would use simulated pulses with exactly known DM and scattering to map where the min-entropy estimator breaks down."],"forward_implications":["Dispersion measurement becomes a template-free search: the DM that concentrates the pulse profile into the fewest flux bins is selected without assuming any pulse shape, sidestepping the fitting bias the paper identifies.","Because the entropy scan is computed separately on each pulse, per-pulse DMs and error bars follow naturally, allowing studies of DM variability without first building a high-S/N template.","The method is computationally cheap enough to supply initial DM estimates and pre-input parameters for more detailed single-pulse analysis, as demonstrated by the coarse-to-fine search in Section 2.2.","On the 48-pulsar FAST sample, a correct minimum-entropy rule predicts that entropy DMs will continue to agree with independently measured DMs, with residual outliers tracing reference-systematic differences rather than method failure."],"supporting_citations":[{"why":"Supplies the definition of information entropy on which the minimum-entropy indicator is built.","marker":"Shannon (1948)"},{"why":"Compiles the ATNF pulsar catalogue used as the latest-reference DM comparison in Figure 4.","marker":"Manchester et al. (2005)"},{"why":"Provides the latest reference DMs for most of the 48 pulsars in Table 1.","marker":"Wang et al. (2023)"},{"why":"Provides the latest reference DMs for the remaining pulsars in Table 1.","marker":"Deneva et al. (2024)"},{"why":"Old reference whose DM agrees with the entropy DM for the outlier J0628+0909.","marker":"Nice et al. (2013)"},{"why":"Old reference whose DM agrees with the entropy DM for the outlier J1851+1259.","marker":"Bilous et al. (2016)"},{"why":"Old reference whose DM agrees with the entropy DM for the outlier J1859+00.","marker":"McEwen et al. (2020)"},{"why":"Describes the FAST telescope whose observations supply the 48-pulsar validation data.","marker":"Jiang et al. (2020)"}],"fun_headline_variants":["Entropy minimum finds pulsar dispersion measures","No fits: minimum entropy scores pulsar DMs","48 pulsars, one entropy rule: DM from least disorder","Minimum Shannon entropy picks true pulse profile","Entropy scan matches known DMs for 45 of 48 pulsars"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The method computes entropy only from data points that survive a 3-sigma cut and a three-segment interference-removal routine, on the assumption that the true pulse stays intact through both; if part of a real pulse lies below the threshold or gets clipped as interference, the entropy minimum can shift to the wrong dispersion measure.","fun_headline_variants_meta":{"raw":{"variants":["Entropy minimum finds pulsar dispersion measures","No fits: minimum entropy scores pulsar DMs","48 pulsars, one entropy rule: DM from least disorder","Minimum Shannon entropy picks true pulse profile","Entropy scan matches known DMs for 45 of 48 pulsars"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000262,"raw_usage":{"total_tokens":1631,"prompt_tokens":1011,"completion_tokens":620,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":627,"completion_tokens_details":{"reasoning_tokens":542}},"tokens_in":627,"tokens_out":620,"duration_ms":5945,"temperature":1.0,"reasoning_tokens":542,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-12T13:37:15.229940+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Inject a synthetic dispersed pulse with a known DM whose true profile has a low-amplitude broad component below the 3-sigma cutoff: if the entropy-DM minimum moves away from the injected DM by more than the quoted error when that weak component is present, and does not move when the component is removed, the preprocessing assumption is falsified.","supporting_citations":[],"review_version":1}