{"id":"71c1145b-ffcb-4751-92c0-b8f5bbab4824","arxiv_id":"2608.07791","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"Splashback galaxies are only a few percent of the population but account for a large part of the galaxy assembly bias signal in simulations; reclassifying them as satellites barely changes the signal.","lead":"This paper studies galaxies that were once inside bigger galaxies and then escaped, called splashback galaxies, in a large computer simulation of the universe. It finds that removing these rare galaxies sharply reduces the measured galaxy assembly bias signal, a subtle extra clustering caused by how dark matter halos formed.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The central attribution of the GAB reduction to splashback-specific physics is not yet secured: no control removes an equally selected non-splashback population to show the effect is not generic to low-mass, dense-environment centrals.","rationale":"The paper is a careful, well-structured empirical study, and the TNG300 check is a genuine independent robustness test. The reader's weakest-assumption identification—that the fly-by definition may misclassify the physically relevant population—is plausible, but I see the more load-bearing gap as one of experimental design rather than object classification. Even with a perfect splashback catalog, the removal experiment would not establish splashback-specific causality unless a matched placebo removal of non-splashback centrals is shown to leave the GAB signal intact. The reported 30%/60% reductions could in principle be reproduced by removing any small set of low-mass centrals living in dense, high-concentration environments, because those are precisely the objects that occupancy-variation and assembly-bias signals are most sensitive to. The paper's own conclusion softens the causal language by calling splashbacks 'tracers' of GAB, which is honest but also exposes the underdetermination. The concrete test I propose would settle this: if the placebo removal mimics the signal reduction, the headline should be reframed as an environmental-selection effect; if it does not, the splashback-specific attribution is strongly supported. Because the current CONDITIONAL verdict already reflects the need for additional checks, my stress-test reading leaves that verdict unchanged while sharpening what the decisive check should be.","tokens_in":22527,"tokens_out":4185,"duration_ms":43615,"concrete_test":"Build a placebo sample from the same n = 0.01 h−3 Mpc−3 catalog: for each splashback central, select a non-splashback central in the same narrow halo-mass, concentration, and environment bins (e.g., same 0.1 dex M_h bin, same concentration tercile, same δ_1.25 quartile), and remove the placebo centrals instead of the splashbacks. Recompute the Fig. 4 correlation-function ratios (full and central-only) and the Fig. 3 occupancy variations, using at least 100 random matchings to bracket the placebo dispersion. If the placebo reproduces the ~30% and ~60% reductions, the splashback-specific identification is not the driver; if the reduction is significantly weaker, the attribution is supported.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The paper's central quantitative claim—removing about 3–4% of galaxies identified as splashbacks reduces GAB by roughly 30% for the full sample and 60% for centrals (§4.3, Fig. 4)—is an attribution claim, but the experiment has no placebo arm. Splashbacks are removed because they preferentially occupy low-mass, high-concentration halos in dense environments (§4.1, Fig. 2). Removing any comparably selected population could reduce the signal by a similar amount, even if fly-by history were irrelevant to assembly bias. The only control reported is raising the stellar-mass threshold to match the reduced number density (§4.3), which checks the overall density dependence but not the selective removal of low-mass, dense-environment, high-concentration centrals. The same gap affects the occupancy-variation 'tail' claim in §4.2: eliminating the tail after removing splashbacks shows that the tail is populated by splashbacks, but not that splashback-specific physics, rather than the shared environmental properties, generates the tail. The paper's own §5 caveat—'there may be no fundamental difference between true splashbacks and galaxies that remained in the outskirts' and 'splashbacks are not necessarily the origin of GAB, but rather tracers of it'—points to exactly this unresolved issue. The reassignment experiment does not close the gap: since galaxy positions are unchanged, it only perturbs the shuffling procedure, not the physical selection. Without a control sample matched on the selection variables, the headline percentages do not uniquely measure the splashback contribution.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"This paper investigates the role of splashback galaxies—defined as present-day central galaxies that were formerly satellites of a more massive host—in galaxy assembly bias (GAB). Using the Guo et al. (2011) semi-analytic model applied to the Millennium Simulation, and a qualitative check with TNG300, the authors construct three stellar-mass selected samples and measure halo occupation variations with concentration and environment, as well as the GAB amplitude from the ratio of the galaxy correlation function to a shuffled sample. They compare the original samples with samples in which splashbacks are removed and samples in which splashbacks are reclassified as satellites of their former hosts. The main findings are that splashbacks preferentially occupy low-mass, highly concentrated halos in dense environments with relatively high stellar-to-halo mass ratios; that they account for the low-mass tails in the occupancy variations for concentrated and dense-environment halos; and that removing splashbacks reduces the GAB signal by roughly 30% for the full sample and about 60% for centrals at n=0.01 h^3 Mpc^-3, while reassigning them leaves the amplitude largely unchanged but shifts the transition scale.","tokens_in":22793,"tokens_out":4682,"duration_ms":44161,"significance":"If the causal attribution holds, this is a valuable result: it identifies a small, individually identifiable galaxy population that accounts for a substantial fraction of GAB in simulations, with direct implications for HOD modeling and for interpreting observations of assembly bias. The study is methodologically transparent, uses publicly available simulations, contains no fitted parameters, and includes a cross-check with a hydrodynamical simulation. These are genuine strengths. However, the central quantitative claim—that splashbacks are largely responsible for the measured reduction—is not yet secured, because the splashback selection shares the very secondary properties (low mass, high concentration, dense environment) that generate the GAB signal. The manuscript's own Section 5 caveat that splashbacks may be tracers rather than the origin of GAB is an appropriate framing, but the abstract and Section 4.3 currently state the stronger causal claim without a placebo control. The recommendation below is therefore major revision rather than acceptance.","major_comments":[{"comment":"The headline claim that removing about 3-4% of galaxies reduces GAB by roughly 30% for the full sample and 60% for centrals is an attribution claim, but the experiment has no placebo arm. Splashbacks are removed because they preferentially occupy low-mass, high-concentration halos in dense environments (Section 4.1, Figure 2), which are exactly the properties that generate the GAB signal. Removing any comparably selected population of non-splashback centrals, matched in halo mass, concentration, environment, and stellar mass, could produce a similar reduction even if fly-by history were irrelevant. The only reported control, raising the stellar-mass threshold to match the reduced number density (Section 4.3), checks the overall density dependence but not the selective removal of low-mass, dense-environment, high-concentration centrals. The authors should add a matched placebo removal, or otherwise demonstrate that the reduction is specifically associated with the splashback fly-by history rather than with the shared environmental and structural properties.","section":"Section 4.3, Figure 4"},{"comment":"The percentage reductions (roughly 30% and 60%) are central to the abstract and conclusions, but they are quoted without propagated uncertainties. The shaded regions in Figure 4 reflect scatter among shuffled samples, and jackknife errors are stated to be subdominant without being shown or quantified. The authors should report uncertainties on the large-scale ratio and propagate them to the reduction percentages. The TNG300 confirmation in Appendix A is explicitly qualitative and, given its smaller volume and larger uncertainties, should not be used to support the specific percentages.","section":"Section 4.3, Figure 4"},{"comment":"The reassignment experiment does not close the attribution gap. Since galaxy positions are unchanged, reassignment changes only the shuffling procedure, not the physical selection of the splashback population, as the paper itself notes in Section 3.2. The green lines in Figure 4 therefore demonstrate insensitivity to halo classification, not that fly-by history is the causal ingredient. The manuscript's Section 5 statement that 'there may be no fundamental difference between true splashbacks and galaxies that remained in the outskirts' and that splashbacks are 'tracers rather than the origin' points directly to this unresolved issue. Unless the placebo test in the first comment is added, the causal language should be softened to the tracer interpretation throughout the abstract and results.","section":"Section 3.1 and Section 4.3"},{"comment":"The claim that removing or reassigning splashbacks 'eliminates' the low-mass tail of the occupancy variation for concentrated halos and dense environments is presented without statistical significance. The effects are described as 'subtle but yet distinct', but no error bars or significance estimates are shown for the halo occupation functions in Figure 3. Since this claim is load-bearing for the conclusion that splashbacks are largely responsible for the occupancy-variation tails, the authors should quantify the significance of the tail reduction, for example with bootstrap resampling of halos or repeated shuffling.","section":"Section 4.2, Figure 3"}],"minor_comments":[{"comment":"There are several typos: 'splasback' in the abstract, 'Millenium' in the caption of Figure 3, and 'splasbacks' in the caption of Figure 4; these should be corrected.","section":"Abstract and captions"},{"comment":"The text of Appendix A states that the TNG300 shaded regions come from 20 randomly shuffled samples, while the caption of Figure 5 says 10; this inconsistency should be reconciled.","section":"Appendix A and Figure 5"},{"comment":"The footnote that galaxy numbers are quoted in units of thousands is easy to misread; consider writing the column header explicitly as 'Ntot (10^3 galaxies)' and similarly for the other counts.","section":"Table 1"},{"comment":"In Figure 3, the dashed lines are labeled 'Reassigned splashbacks', but the text also uses this figure to infer the removal case by reading the low-mass central occupation. A sentence clarifying which curves correspond to the removal scenario would improve readability.","section":"Section 4.2"},{"comment":"The shuffling procedure is described only verbally; specifying the exact halo-mass binning (width and range) and the treatment of halos with no assigned galaxies would aid reproducibility, even though the authors state that results are insensitive to binning.","section":"Section 3.2"}],"recommendation":"major_revision","confidential_remarks":"The stress-test concern is real and is the main reason I recommend major revision rather than acceptance. The missing placebo arm directly affects the paper's central attribution claim, but it is fixable within the scope of the manuscript: add a matched removal of non-splashback centrals and report uncertainties on the percentage reductions. If the authors prefer not to add the placebo analysis, the paper would still be defensible if the abstract and conclusions are consistently softened to the tracer interpretation that the authors themselves articulate in Section 5. The reassignment experiment and the TNG300 qualitative check are useful supporting material but do not resolve the attribution issue."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Colleague,\n\nThis paper directly quantifies what splashback galaxies contribute to galaxy assembly bias in the SAM/Millennium setup. That is new: prior work linked splashback halos to halo assembly bias, but this is the first to isolate the galaxy population and show that removing roughly 3-4% of galaxies identified as splashbacks cuts the GAB signal by about a third for the full sample and roughly 60% for centrals at n=0.01. The reassignment test is a nice twist: since positions are unchanged, it speaks to halo-finder sensitivity rather than physical selection. The TNG300 check, though qualitative, is a genuine robustness test. The paper is clear, the methods are standard, the data are public, and the authors cite the relevant prior work.\n\nThe soft spot is the one the stress-test flags. The headline percentages are an attribution claim without a placebo arm. Splashbacks preferentially sit in low-mass, high-concentration halos in dense environments. Removing any comparably selected population might reduce the signal just as much, even if fly-by history were irrelevant. The only control is raising the stellar-mass threshold to match number density, which checks density dependence but not the selective removal of low-mass, dense-environment, high-concentration centrals. The occupancy-variation tail argument has the same gap: it shows the tail is populated by splashbacks, not that splashback-specific physics generates it. The authors themselves concede this in Section 5, saying splashbacks are not necessarily the origin of GAB but rather tracers of it, so the gap is acknowledged. The reassignment experiment cannot close it, since positions are unchanged. I also note the headline reduction percentages have no error bars; the shaded regions are shuffle-to-shuffle uncertainty, and jackknife errors are stated to be subdominant but not shown. TNG300 is qualitative only.\n\nNone of this kills the paper. The descriptive result, that a small, cleanly identifiable population carries a disproportionate share of GAB in this SAM, is solid and useful for HOD modeling. What is not secured is the causal story. A referee should ask for a control sample matched on the selection variables, e.g., removing non-splashback centrals in the same low-mass, high-concentration, dense-environment region, and for error bars on the percentage reductions.\n\nI would send this to review. It is a serious empirical contribution that will be read by the galaxy-halo connection crowd, even if the mechanism question remains open.","headline":"A solid, useful empirical study showing splashback galaxies carry a disproportionate share of galaxy assembly bias, with a genuine causal gap that the authors themselves acknowledge and a referee can close with a control sample.","tokens_in":23331,"tokens_out":2139,"would_cite":true,"duration_ms":21074,"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":"This paper claims that splashback galaxies account for a large share of galaxy assembly bias in simulations, with removal tests cutting the full-sample signal by about a third and the central-only signal by about 60%.","keywords":["galaxy assembly bias","splashback galaxies","halo occupation function","occupancy variations","halo concentration","large-scale environment","semi-analytic galaxy formation","cosmological simulations"],"falsifier":"Repeat the removal test using splashbacks identified by a different scheme, such as halos between a massive host's virial radius and its splashback radius, or halos found by a different halo finder; if removing that population does not shrink the shuffled-clustering ratio by roughly 30% for the full sample and 60% for centrals, the claimed role of splashback galaxies is an artifact of the fly-by definition rather than a physical driver.","tokens_in":22341,"feed_emoji":"🌌","tokens_out":6829,"duration_ms":56557,"temperature":0.7,"pith_summary":"This paper sets out to determine whether splashback galaxies, defined as present-day central galaxies that were once satellites of a more massive host, are a main physical source of galaxy assembly bias. Using a semi-analytic galaxy formation model on a large dark-matter simulation, the authors identify splashbacks from merger-tree histories and then either remove them from stellar-mass-selected samples or reassign them as satellites of their former hosts. They find that splashbacks are few, about 2–4% of all galaxies and up to roughly 8% of centrals, yet they sit in low-mass, highly concentrated halos in dense environments with elevated stellar-to-halo mass ratios. Removing them eliminates most of the low-mass tail of the occupancy variation for those halo populations and reduces the galaxy assembly bias signal by roughly 30% for the full sample and about 60% for centrals alone, while reassigning them leaves the signal amplitude nearly unchanged.","feed_headline":"Splashback galaxies account for 30% of galaxy assembly bias","feed_subtitle":"Just 3% of galaxies, centrals that once fell through bigger halos, produce most of the low-mass clustering excess.","key_machinery":"The central object is the splashback galaxy, defined by its fly-by merger-tree history: a present-day central galaxy that was previously a satellite of a more massive host halo. The second piece of machinery is the shuffling test for galaxy assembly bias, in which galaxies are randomly reassigned to halos of the same mass, erasing secondary-property dependence, and the ratio of the original correlation function to the shuffled one isolates the assembly-bias signal. Comparing the original, splashback-removed, and splashback-reassigned samples under this test is what carries the argument.","core_discovery":"The paper argues that splashback galaxies, rather than generic halo age or concentration effects alone, are largely responsible for the low-mass tail of halo occupancy variation seen for highly concentrated halos and for halos in dense environments, and that they account for a substantial fraction of the measured galaxy assembly bias. The load-bearing comparison is between a shuffled sample that destroys assembly bias and samples in which splashbacks are removed or reclassified; the central claim is that removing about 3% of the galaxies reduces the large-scale correlation-function excess by roughly a third, and removing about 4% of centrals reduces it by about 60%. Because reassignment rather than removal has little effect on the amplitude, the paper concludes that the effect does not depend on whether a halo finder labels these objects as centrals or as satellites; what matters is the environmentally truncated growth the splashback label traces. The same pattern appears in a hydrodynamical simulation, so the finding is not tied to the particular semi-analytic model.","pith_inferences":["A natural extension, not tested here, would be to identify splashbacks by splashback radius rather than fly-by histories; if the same ~30% reduction appears, the mechanism is robust to the definition rather than an artifact of merger-tree tracking.","The paper's emphasis on removal versus reassignment suggests that HOD-style models could treat splashbacks either way and still reproduce clustering; the testable step is to see whether emulators built on either convention match small-scale clustering equally well.","Because the splashback contribution grows at lower stellar-mass thresholds, deeper surveys reaching fainter galaxies should expect a stronger assembly-bias contamination; this is an inference from the paper's density dependence, not a result it derives."],"forward_implications":["If the claim holds, halo occupation models that ignore splashbacks will overestimate the role of halo concentration and environment in the low-mass regime, because much of that signal is carried by a few percent of objects.","Because reassigning splashbacks does not change the signal amplitude, galaxy assembly bias measurements in simulations do not depend on whether a halo finder classifies these objects as centrals or satellites.","Removal of splashbacks gives the largest GAB reduction in the paper's lower stellar-mass-threshold samples, so deeper galaxy samples should show an even stronger splashback contribution to assembly bias.","The residual assembly bias after removal indicates that roughly two-thirds of the full-sample signal, and a measurable part of the central-only excess, comes from other environmental and assembly effects, so splashbacks are a major but not the sole driver.","The result is qualitatively reproduced in a hydrodynamical simulation, suggesting the splashback contribution is a feature of the underlying halo growth in dense environments rather than a quirk of one galaxy formation model."],"supporting_citations":[{"why":"Supplies the Millennium N-body simulation, halo and subhalo catalogs, and merger trees used for the main analysis.","marker":"Springel et al. 2005"},{"why":"Provides the semi-analytic galaxy formation model that produces the stellar-mass-selected galaxy samples.","marker":"Guo et al. 2011"},{"why":"Introduces the shuffling methodology that the paper uses to measure galaxy assembly bias from correlation-function ratios.","marker":"Croton et al. 2007"},{"why":"Establishes the occupancy-variation framework and the concentration and environment measures used here, and provides the baseline GAB comparison.","marker":"Zehavi et al. 2018"},{"why":"Shows that splashback halos account for a significant part of halo assembly bias, the dark-matter-level result this paper extends to galaxies.","marker":"Mansfield & Kravtsov 2020"},{"why":"Suggests splashbacks may be mislabeled satellites of their former host halos, motivating the reassignment treatment.","marker":"Diemer 2021"},{"why":"Finds that splashbacks dominate the low-mass tail of the central galaxy occupation function, the trend this paper quantifies for GAB.","marker":"Contreras et al. 2013"},{"why":"Shows that ejected satellites have high stellar-to-halo mass ratios, supporting the paper's interpretation of splashback stellar masses.","marker":"Wetzel et al. 2014"},{"why":"Provides the TNG300 hydrodynamical simulation used for the robustness check in Appendix A.","marker":"Nelson et al. 2019"}],"fun_headline_variants":["Splashback galaxies drive most of assembly bias","Removing 3% of galaxies kills 30% of assembly bias","Splashbacks, not halo age, cause low-mass clustering excess","Rare splashback galaxies explain galaxy clustering excess","Assembly bias traced to galaxies that fell through bigger halos"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The fly-by definition of a splashback galaxy, a present-day central that was once a satellite of a currently more massive host, correctly isolates the physically relevant population, so that the measured reduction in assembly bias comes from these objects rather than from merger-tree errors or from excluding major mergers.","fun_headline_variants_meta":{"raw":{"variants":["Splashback galaxies drive most of assembly bias","Removing 3% of galaxies kills 30% of assembly bias","Splashbacks, not halo age, cause low-mass clustering excess","Rare splashback galaxies explain galaxy clustering excess","Assembly bias traced to galaxies that fell through bigger halos"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000374,"raw_usage":{"total_tokens":2030,"prompt_tokens":1009,"completion_tokens":1021,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":625,"completion_tokens_details":{"reasoning_tokens":939}},"tokens_in":625,"tokens_out":1021,"duration_ms":6975,"temperature":1.0,"reasoning_tokens":939,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-11T04:12:56.965261+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Repeat the removal test using splashbacks identified by a different scheme, such as halos between a massive host's virial radius and its splashback radius, or halos found by a different halo finder; if removing that population does not shrink the shuffled-clustering ratio by roughly 30% for the full sample and 60% for centrals, the claimed role of splashback galaxies is an artifact of the fly-by definition rather than a physical driver.","supporting_citations":[],"review_version":1}