{"id":"3ea9ffb0-c738-490d-9e41-314c7d864638","arxiv_id":"2611.17009","paper_version":1,"verdict":"CONDITIONAL","confidence":"HIGH","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":6,"one_line_summary":"A tracer-centred density-split clustering model is developed that directly models redshift-space distortions and shows that the cosmological information gain over standard two-point statistics comes from the density PDF's incorporation of small-scale modes.","lead":"This paper develops a new modelling framework for density-split clustering in redshift space, using a tracer-centred density PDF instead of the usual volume-weighted one. The approach avoids the problem of voxel non-conservation between real and redshift space and enables direct modelling of redshift-space distortions without reconstruction.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek/deepseek-v4-flash","headline":"The model's central claim that information gain comes from the density PDF incorporating small-scale modes is supported, but the empirical model's reliance on 5 nuisance parameters per bin means the Fisher forecast's apparent gain is driven by fixing those parameters; when they are marginalized…","rationale":"The reader's weakest_assumption correctly identifies the scatter in the real-to-redshift-space density mapping (Eq. 6) as a concern. However, the paper acknowledges this scatter (Fig. B1) and shows the model works at the few-percent level for the fiducial scale. The more load-bearing concern is the empirical nature of the model and the fragility of the Fisher forecast's information gain to marginalization over nuisance parameters and to realistic effects like FoG and unknown bias. The reader's verdict is CONDITIONAL, which I agree with, but my reasoning emphasizes the empirical model's flexibility as the primary risk, rather than the scatter in the mapping. The paper's own limitations (b174, b175, b180, b281) support this. The concrete test I propose directly tests robustness under more realistic conditions, which is the key unresolved issue.","tokens_in":29800,"tokens_out":737,"duration_ms":8747,"concrete_test":"Re-run the Fisher forecast of Fig. 11 with the same k-range for DS and Global (as in Fig. B4) but also marginalize over b_g (as in Appendix C) and include a simple FoG damping term (e.g., exp(-k^2 sigma_v^2) with sigma_v ~ 4 h^-1 Mpc) in both DS and Global models. If the combination gain disappears or the degeneracy direction aligns with Global, then the claimed information gain from the PDF is not robust to realistic RSD modelling.","verdict_should_be":"CONDITIONAL","load_bearing_attack":"The paper's strongest claim (b279) is that the extra cosmological information in DS clustering comes from the density PDF via b_DS, which depends on sigma^2(R) and xi_bar(R) and thus incorporates high-k modes. However, the Fisher forecast that supports this claim (Fig. 11) shows that when the 5 nuisance parameters per bin (A0, A2, n0, n2, alpha) are marginalized over, DS clustering alone is substantially weaker than the Global power spectrum (Panel A). The combination of DS+Global gives tighter constraints than Global alone only because the degeneracy directions differ. But this difference in degeneracy direction is itself contingent on the assumption that the nuisance parameters are well-behaved and that the model for b_DS is accurate. The paper acknowledges (b174, b175) that the P_NLC term is not from first principles and must be marginalized, which degrades constraints. The key load-bearing concern is that the claimed information gain from the PDF is not robustly separable from the empirical model's flexibility: the gain appears only when nuisance parameters are fixed (Panel B) or when combining with Global, and the combination gain relies on the specific degeneracy direction which may change with more realistic physics (FoG, non-linear bias, survey geometry). The paper's own Appendix C shows that marginalizing over linear bias b_g further weakens constraints. Thus, the central claim is not falsified but is not yet demonstrated to be robust under realistic conditions.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek/deepseek-v4-flash","summary":"This paper develops a modelling framework for density-split (DS) clustering in redshift space, centring the split on a tracer-centred density PDF rather than the standard volume-weighted counts-in-cells PDF. The authors derive a linear bias b_DS associated with each density threshold in redshift space (Section 3, Eq. 7), connect the DS power spectra to this bias through an empirical halo-model-inspired decomposition (monopole and quadrupole, Eqs. 8--14), and validate the model against approximate lognormal simulations at z=2, reporting few-percent accuracy in monopole and quadrupole up to k~1 h/Mpc. They then perform Fisher forecasts (Section 5, Fig. 11) for Omega_m and sigma_8, comparing DS clustering, the global power spectrum, and their combination, under several treatments of nuisance parameters and of b_DS. The central claim (b279) is that the extra cosmological information in DS clustering comes primarily from b_DS, which depends on the tracer-centred PDF and thereby on high-k modes through sigma^2(R) and xi_bar(R), even when the clustering data vector is cut at lower k. Additional support is provided by a Gaussian-field test at z=10 (Fig. 12) and by an appendix (Appendix C) that marginalises over a free linear bias b_g, showing qualitative persistence of the information through b_DS but with degraded overall constraints.","tokens_in":30379,"tokens_out":1700,"duration_ms":17389,"significance":"If the claimed mechanism is correct, the paper provides a valuable conceptual clarification: DS clustering gains information not solely from non-Gaussianity of the PDF, but from the fact that the DS bias couples to modes below the smoothing scale even when the measured power spectrum is truncated. The derivation of b_DS from a tracer-centred PDF in redshift space is a genuine methodological step forward, and the validation using a controlled simulation setup is a useful benchmark. The paper is careful to show both optimistic (nuisance parameters fixed) and conservative (nuisance parameters marginalised) cases, and it explicitly demonstrates that the information gain disappears when b_DS is held fixed (Fig. 11C,D; Fig. B4). The Gaussian-field test strengthens the interpretation that the effect is mode reconstruction through the PDF rather than non-Gaussianity per se. The main weakness is that the model is empirical and the forecast gain in the realistic marginalised case is modest, appearing only when DS is combined with the global spectrum and when different k-cuts are adopted.","major_comments":[{"comment":"The claim that DS clustering provides tighter cosmological constraints than the global two-point statistic is qualified by the paper itself: when the five nuisance parameters per bin are marginalised (Panel A), DS clustering alone is substantially weaker than Global, and the combined gain relies on different degeneracy directions and different k-cuts (Fig. B4 shows no gain with equal k-cuts). The abstract's phrasing 'can provide tighter constraints' is too strong without stating the conditions under which this holds. I recommend that the abstract and conclusions explicitly state that the gain is demonstrated only when nuisance parameters are fixed or when DS is combined with Global using a larger k-range for DS, and that the marginalised-alone case shows a significant degradation.","section":"Section 5, Fig. 11 and Fig. B4","authors_acknowledge":"The present model includes five nuisance parameters per den- sity bin, which weakens the constraining power on cosmological parameters once marginalised over.","authors_acknowledge_p":0.75},{"comment":"The mapping from real-space to redshift-space tracer-centred PDF relies on Eq. (5), which ignores scatter around the mean linear relation. The paper acknowledges this scatter in Fig. B1, noting it increases at smaller R, but the forecast and the DS-bias computation use the fiducial R=5 h^-1 Mpc, where scatter is present. This is a load-bearing approximation: if the scatter were modelled, b_DS could differ and the Fisher-information gain through b_DS might change. The authors should either quantify the impact of this scatter on b_DS and the forecasts, or explicitly state that the current forecasts assume a linear mapping that is known to be imperfect at the fiducial smoothing scale.","section":"Section 3.3, Eq. (6) and Fig. B1","authors_acknowledge":null,"authors_acknowledge_p":null},{"comment":"The empirical P_NLC term is not derived from first principles; the paper says so (b174) and that marginalisation will degrade constraints. This is acceptable for a framework paper, but the Fisher forecast's treatment of these terms matters. In the marginalised cases, the information gain from DS alone is small, and the combination gain depends on the degeneracy direction, which could be sensitive to the assumed form of P_NLC. I suggest adding a brief discussion or test of how the conclusions change if the nuisance parameters are allowed to have different functional forms (e.g., including a k-dependence in the stochastic term or a separate FoG-like damping). Without such a test, the robustness of the claimed information gain is not fully established.","section":"Section 4, Eqs. (8)--(14)","authors_acknowledge":"Athigh-𝑘(𝑘∼1ℎMpc −1),theclusteringisdominatedbythe𝑃 NLCℓ term.Untilthistermiswellunderstoodfromfirstprinciples,wewill have to marginalise over the nuisance parameters we use to describe it (𝐴ℓ,𝑛ℓ,𝛼and the choice of smoothing window).","authors_acknowledge_p":0.61},{"comment":"When the linear bias b_g is free, the constraints on sigma_8 degrade considerably, and the qualitative gain of DS over Global is reduced. The paper acknowledges this in Appendix C, but the main-text abstract and conclusion do not mention this important caveat. I recommend that the abstract or at least the conclusions mention that the information gain through b_DS is demonstrated for fixed linear bias, and that marginalising over b_g weakens the constraints but does not erase the different degeneracy direction between DS and Global.","section":"Appendix C","authors_acknowledge":"APPENDIX C: IMPACT OF MARGINALISING OVER THE LINEAR BIAS IN THE FISHER FORECAST In the forecasts shown here, we present the conservative case where the observed𝑃 ℓ(𝑘)are fixed, and the tracer-centred density split valuesarealsofixedtothesameconstantfractionsofthefullsample (e.g. those used in the main text).","authors_acknowledge_p":0.66}],"minor_comments":[{"comment":"The abstract says 'uniformly random' (probably a typo) and the slogan 'doughnut-like structures' refers to the Islands; consider clarifying that these are defined through thresholded smoothed fields, not actual physical objects.","section":"Abstract and Section 6","authors_acknowledge":null,"authors_acknowledge_p":null},{"comment":"The caption of Fig. 2 mentions dashed curves for the model, but the figure itself is not shown in the text; check that the PDF model prediction is actually presented or referenced in the appendix.","section":"Section 2.2, Fig. 2","authors_acknowledge":null,"authors_acknowledge_p":null},{"comment":"The variance constraint used to set C is sensible but the assumption that P_s0(k) reproduces the variance completely is not tested. A brief sentence on how well Eq. (15) is satisfied by the measured power spectra would help.","section":"Section 4, Eq. (15)","authors_acknowledge":null,"authors_acknowledge_p":null},{"comment":"The Fisher forecast uses linear k-bins with a fixed range; it would be helpful to specify the exact bin width and the number of bins used in the data vector to allow reproducibility.","section":"Section 5, Fisher setup","authors_acknowledge":null,"authors_acknowledge_p":null},{"comment":"In several places, \\hat{\\delta}_R is used without a clear definition of the hat; I assume it denotes the smoothed density contrast, but the notation is inconsistent (e.g., \\hat{\\delta}_R vs \\delta_R). Please unify.","section":"Notation","authors_acknowledge":null,"authors_acknowledge_p":null},{"comment":"The text refers to Fig. B2 and B3, but the captions do not list the parameter values; it would help to include the numerical values of the nuisance parameters in the caption or in the text.","section":"Appendix B figure labels","authors_acknowledge":null,"authors_acknowledge_p":null}],"recommendation":"major_revision"},"author_rebuttal":null,"desk_editor":null,"pith_extraction":{"msc":[],"pacs":["98.80.-k","98.65.-r"],"model":"deepseek/deepseek-v4-flash","headline":"Density-split clustering gains cosmological information through the density PDF, not just the power spectrum","keywords":["density-split clustering","redshift-space distortions","tracer-centred PDF","beyond two-point statistics","cosmological information","large-scale structure","marked statistics","Fisher forecast"],"falsifier":"If a simulation at $z=2$ with a more realistic galaxy formation model (e.g., hydrodynamical simulation or halo occupation distribution) shows that the tracer-centred density PDF in redshift space deviates from the linear mapping of Eq. 6 by more than the few-percent level claimed, or if the DS bias $b_{\\rm DS}$ computed from the PDF does not match the measured large-scale amplitude of DS clustering to within the stated accuracy, then the model's predictive power would be falsified.","tokens_in":31278,"feed_emoji":"🔭","tokens_out":1943,"duration_ms":12123,"temperature":0.7,"pith_summary":"This paper develops a framework to model density-split (DS) clustering directly in redshift space, solving a key problem that prevented its use with real data. The authors show that DS clustering extracts more cosmological information than standard two-point statistics because the DS bias $b_{\\rm DS}$ depends on the tracer-centred density probability distribution function (PDF), which itself depends on the matter power spectrum up to very high $k$-modes through the variance $\\sigma^2(R)$ and mean correlation $\\bar{\\xi}(R)$. This means that even when the clustering analysis is cut at a relatively large scale, the amplitude of DS clustering on large scales still encodes information from small-scale modes that are inaccessible to the standard power spectrum. The paper provides a step-by-step procedure for computing the tracer-centred PDF in redshift space, deriving the linear bias associated with each density bin, and constructing an empirical halo-model-like description of the DS power spectra. Fisher forecasts confirm that combining DS clustering with the global power spectrum can tighten constraints on $\\Omega_m$ and $\\sigma_8$ due to their different degeneracy directions, and that the extra information comes specifically from the cosmological dependence of $b_{\\rm DS}$.","feed_headline":"How density-split clustering beats the power spectrum","feed_subtitle":"The gain comes from the density PDF, which captures small-scale modes even when large-scale clustering is measured.","key_machinery":"The central object is the tracer-centred density probability distribution function (PDF) $P_{\\rm TC}(\\hat\\delta^s_R)$, which conditions the density field on the locations of tracers (galaxies/halos) rather than on random volume elements. This PDF is computed by starting from the real-space volume-weighted PDF of the matter density field (Eq. 1, from Uhlemann et al. 2016), converting it to the tracer-centred version (Eq. 3, derived in Appendix A), and then mapping it to redshift space using a linear relation (Eq. 6) that averages over the angle-dependent Kaiser factor. From this PDF, the density-split bias $b_{\\rm DS}$ for each density bin is obtained via Eq. 7, which is the key quantity that carries the extra cosmological information. The DS power spectra are then modelled via an empirical halo-model-like decomposition (Eqs. 8-14) that separates a linear two-Island term (proportional to $b_{\\rm DS}^2$ times the Kaiser-boosted matter power spectrum), a non-linear correction term $P^{\\rm NLC}_\\ell$, and a stochastic term $P^{\\rm stoch}$.","core_discovery":"The paper establishes that the gain in cosmological information from density-split (DS) clustering over standard two-point statistics arises primarily from the density-split bias $b_{\\rm DS}$, which depends on the tracer-centred density PDF. This PDF incorporates small-scale perturbation modes below the smoothing scale through the variance $\\sigma^2(R)$ and mean correlation $\\bar{\\xi}(R)$. Consequently, even when the DS clustering analysis is restricted to large scales (low $k$), the amplitude of the DS power spectra retains information from high-$k$ modes that are cut off in the global power spectrum. The authors develop a complete modelling framework that works directly in redshift space by switching from volume-weighted counts-in-cells to a tracer-centred density PDF, thereby avoiding the problem of non-conservation of density-split centres between real and redshift space. They derive the linear bias for each density bin in redshift space, construct an empirical halo-model-like description of the DS monopole and quadrupole power spectra, and validate the model against approximate lognormal simulations at $z=2$, achieving few-percent accuracy up to $k\\sim 1\\,h\\,{\\rm Mpc}^{-1}$.","pith_inferences":["The paper's demonstration that DS information comes from the PDF implies that any survey analysis that measures the galaxy density PDF (e.g., via counts-in-cells) already contains some of the same information, and that combining the PDF with clustering may yield even tighter constraints than DS clustering alone.","The empirical halo-model-like parameterisation of the DS power spectra, while effective, leaves the physical origin of the $P^{\\rm NLC}_\\ell$ term unclear; a first-principles derivation of this term (e.g., from perturbation theory or a proper one-Island model) would eliminate the need for nuisance parameters and unlock the full constraining power of DS clustering.","The model was validated only on idealised lognormal simulations at $z=2$ with no Finger-of-God effect; applying it to realistic N-body simulations and survey data will test whether the few-percent accuracy holds in the presence of non-linear velocities, assembly bias, and complex survey geometry.","The Fisher forecasts assume fixed density bins and perfect knowledge of the linear bias $b_g$; in practice, marginalising over $b_g$ and the bin edges will degrade constraints, though the Appendix shows that DS clustering still provides some advantage when $b_g$ is free."],"forward_implications":["Density-split clustering can be applied directly to galaxy redshift surveys without needing reconstruction or emulation, enabling a new observational probe of cosmology.","Combining DS clustering with the global power spectrum can tighten constraints on $\\Omega_m$ and $\\sigma_8$ because their degeneracy directions differ, even when DS clustering alone is less constraining after marginalising over nuisance parameters.","The extra cosmological information in DS clustering is robustly traced to the density PDF, meaning that any analysis that uses the PDF (e.g., counts-in-cells, marked statistics) can similarly access small-scale modes through large-scale amplitudes.","The model's few-percent accuracy on scales up to $k\\sim 1\\,h\\,{\\rm Mpc}^{-1}$ suggests that DS clustering can extend the range of perturbation theory to smaller scales than the global power spectrum, provided the non-linear term $P^{\\rm NLC}_\\ell$ is well understood.","The framework naturally connects DS clustering to indicator power spectra and marked statistics, unifying several beyond-two-point approaches under a common formalism."],"supporting_citations":[{"why":"Provides the analytic expression for the real-space volume-weighted matter density PDF (Eq. 1) that is the starting point for the tracer-centred PDF model.","marker":"Uhlemann et al. 2016"},{"why":"Derives the density-split bias formula (Eq. 7) and establishes the connection between DS clustering and indicator power spectra.","marker":"Repp & Szapudi 2022"},{"why":"Demonstrate that DS clustering can provide tighter cosmological constraints than the global two-point function, which the present paper confirms and explains.","marker":"Paillas et al. 2021, 2023"},{"why":"First density-split analysis of galaxy clustering using tracer-centred densities, providing the conceptual foundation for the present work.","marker":"Abbas & Sheth 2007"},{"why":"The standard linear RSD model that provides the $K_\\ell(\\beta_g)$ factors used in the DS power spectrum model.","marker":"Kaiser 1987"},{"why":"Provides the large-deviation theory framework for the matter PDF that underpins Eq. 1.","marker":"Codis et al. 2016"}],"fun_headline_variants":[],"cache_read_input_tokens":0,"weakest_assumption_plain":"The mapping from real-space to redshift-space tracer-centred density PDF assumes a linear relation (Eq. 5) with no scatter, but the paper acknowledges (Fig. B1) that there is noticeable scatter that increases as the smoothing scale decreases; for the fiducial smoothing scale of $5\\,h^{-1}\\mathrm{Mpc}$, this scatter could be significant and may degrade the model's accuracy for different cosmologies or smaller scales.","fun_headline_variants_meta":{"raw":{"variants":[]},"model":"deepseek/deepseek-v4-flash","effort":"low","cost_usd":0.000235,"raw_usage":{"cost":0.0002345252,"is_byok":false,"cost_details":{"upstream_inference_cost":0.0002345252,"upstream_inference_prompt_cost":7.92792e-05,"upstream_inference_completions_cost":0.000155246},"total_tokens":1929,"prompt_tokens":1076,"completion_tokens":853,"prompt_tokens_details":{"audio_tokens":0,"video_tokens":0,"cached_tokens":256,"cache_write_tokens":0},"completion_tokens_details":{"audio_tokens":0,"image_tokens":0,"reasoning_tokens":0}},"tokens_in":1076,"tokens_out":853,"cost_source":"openrouter_usage","duration_ms":5642,"image_pages":[],"temperature":1.0,"billed_attempts":1,"reasoning_tokens":0,"transport_provider":"openrouter","cache_read_input_tokens":0,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-11-13T07:32:36.799335+00:00","model_set":{"reader":"deepseek/deepseek-v4-flash"},"pipeline_version":"pith-pipeline@v1.4.1-daily-deepseek-or-medium","input_kind":null,"input_chars":null,"input_total_chars":null,"passes_missing":null,"published_standing":null,"published_confidence":null,"objection_audit":null,"run_record":null,"basis":{"kind":"not_recorded","source":null,"chars_given":null,"chars_total":null},"receipt":null,"falsifier":"If a simulation at $z=2$ with a more realistic galaxy formation model (e.g., hydrodynamical simulation or halo occupation distribution) shows that the tracer-centred density PDF in redshift space deviates from the linear mapping of Eq. 6 by more than the few-percent level claimed, or if the DS bias $b_{\\rm DS}$ computed from the PDF does not match the measured large-scale amplitude of DS clustering to within the stated accuracy, then the model's predictive power would be falsified.","supporting_citations":[],"lane_accuracy":{"status":"not yet measured","lane":{"model":"deepseek/deepseek-v4-flash","input_kind":null},"url":"https://pith.science/accuracy"},"review_version":1}