{"id":"222849eb-cb98-4413-a994-26be6bfc5014","arxiv_id":"2508.06608","paper_version":2,"verdict":"ACCEPT","confidence":"MODERATE","novelty_score":4.0,"correctness_risk":"low","formal_verification":"none","parameter_count":0,"one_line_summary":"FIRE-2's second data release makes 119 cosmological zoom-in galaxy simulations publicly available, including full snapshot histories, halo catalogs, and merger trees.","lead":"This paper announces the second public data release of the FIRE-2 galaxy formation simulations, adding full snapshot histories, new physics variants, and hundreds of simulations to a public archive. It matters because these simulations are a widely used foundation for studying how galaxies form and evolve.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Central claim is an availability claim; the text is internally careful, but the load-bearing premise (files exist on FlatHUB as described) can only be settled by direct inspection of the archive.","rationale":"The paper's job is to document an archive. Internally, the description is mostly coherent: suite counts, physics variants, directory names, and limitations (spurious cosmic-ray heating, missing snapshots, m09 resolution) are disclosed, and the reliance on notes.txt is a reasonable mitigation. DR1's existence at the same host independently supports the claim that FlatHUB is a working distribution channel, so the risk of a nonexistent host is low. The one external premise that cannot be certified by the text is that the actual FlatHUB tree matches the printed manifest. I found two small text/table mismatches—z9 filenames in Section 2.3 vs Table 1, and the 5 SMC/LMC + 4 lower-mass count vs six m11 rows in Table 1—which are the kind of metadata error that would be caught by the same directory audit. They are not sufficient to reject the release, but they are the closest thing to a concrete weakness in the central claim. A direct listing of the archive settles the question. Given the moderate confidence and the low scientific risk of a data-release paper, the reader's ACCEPT verdict remains appropriate; the external check is a verification step, not a ground for changing the verdict.","tokens_in":16913,"tokens_out":10039,"duration_ms":113643,"concrete_test":"Access the FlatHUB endpoint (browser or Globus) and programmatically compare a recursive directory listing of flathub.flatironinstitute.org/fire against Table 1 and Section 2.3: suite names, per-simulation snapshot counts, existence of notes.txt, and presence of tree.hdf5 for the Core merger-tree simulations. In particular, list the z9 directory names to see whether Section 2.3 or Table 1 is the accurate file manifest.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The paper's central claim is not a physics result but a data-availability claim: the stated simulations, snapshot counts, catalogs, and merger trees are publicly readable at flathub.flatironinstitute.org/fire with the directory structure in Table 1. For this to be true, every asserted path and file must exist: 601 snapshots for core Base and Later Reionization simulations, 278 for Massive Halo, 68/42/27 for High Redshift, 61 for DMO/MHD+/CR subsets, plus halo catalogs and tree.hdf5 where claimed. Nothing in the manuscript can establish this; it is an external empirical premise. The paper is transparent about known limitations (e.g., cosmic-ray heating bug, missing snapshots, m09 resolution), which is good, but the exceptions are delegated to notes.txt files that themselves can only be checked after access. There is a minor internal warning: Section 2.3 lists z9 simulations as z9m11a,b,c,d,e,z9m12a while Table 1 lists z9m11a,b,c,z9m12a,b,c, so the prose and table disagree on exact file names in the z=9 suite. Also, Section 2.1 says 5 SMC/LMC + 4 lower-mass galaxies while Table 1 shows six m11 runs under 10^11 Msun, so the categorical counts need reconciliation. These do not by themselves overturn the release, but they reinforce that the text-to-archive mapping is the fragile link.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"This manuscript describes the second public data release (DR2) of the FIRE-2 cosmological zoom-in simulations, hosted at FlatHUB. It enumerates the simulations in the Core, Massive Halo, High Redshift, and Cosmological Box suites; gives snapshot counts and directory names; describes newly released physics variations (dark-matter-only, later reionization, MHD+, cosmic rays); documents halo/galaxy catalogs and merger trees; and provides access, licensing, and citation instructions. The paper is a data-description paper; its central claim is that the stated files exist on flathub.flatironinstitute.org/fire with the described contents and directory structure.","tokens_in":17253,"tokens_out":6250,"duration_ms":70470,"significance":"If the archive matches the description, this release is a substantial community resource: it greatly expands the publicly available FIRE-2 outputs, adds physics variants of the Core suite, includes halo/galaxy catalogs and merger trees for most simulations, and transparently documents known caveats (the spurious cosmic-ray heating bug in Base Physics DR1 runs, missing snapshots, and the lack of AGN feedback in B1/B2/C1/C2). The paper is careful in citing prior FIRE papers and in describing file formats. It does not present new physics or falsifiable predictions; its value is archival and reproducibility-oriented. Because users will rely on Table 1 and the section text to locate specific simulations, internal consistency between the text and the table is load-bearing for the paper's central availability claim.","major_comments":[{"comment":"The prose lists the z=9 suite as z9m11a, z9m11b, z9m11c, z9m11d, z9m11e, z9m12a, while Table 1 lists z9m11a,b,c and z9m12a,b,c. These are different simulation sets. Please reconcile the two lists and verify that the directory names on FlatHUB match the published list, since this is the primary way users will locate the released runs.","section":"Section 2.3 vs. Table 1"},{"comment":"The Abstract and Section 2.1 state that the Core suite comprises 14 Milky Way-mass, 5 SMC/LMC-mass, and 4 lower-mass galaxies. Table 1 lists six m11 simulations at 10^11 Msun (m11b,d,e,h,i,q) and only three clearly lower-mass entries (m09, m10q, m10v). The stated counts do not add up to 23. Please clarify the mass classification used for each simulation or correct the counts in the text/table.","section":"Section 2.1 and Abstract vs. Table 1"},{"comment":"The text says each Dark Matter Only simulation includes a subset of 61 snapshots (every tenth), but later states that \"we include (sub)halo and galaxy catalogs and merger trees generated across all 600 snapshots.\" This is contradictory unless catalogs/trees are stored separately for snapshots whose particle data are not released. Please state explicitly how many snapshots have particle data, how many have catalogs, and whether the 600-snapshot claim refers to a different data product.","section":"Section 2.1.2"}],"minor_comments":[{"comment":"Typo in the suggested citation sentence: \"publi,cly\" should be \"publicly\".","section":"Section 3"},{"comment":"Typo: \"multipling\" should be \"multiplying\".","section":"Section 2.1.2"},{"comment":"The sentence beginning \"Because of their lack of AGN feedback, the primary galaxies in all of the Massive Halo simulations...\" is confusing, since the preceding sentence states that A1, A2, A4, A8 do track SMBH growth. Please rephrase to make clear which simulations lack AGN feedback and which are affected by the overmassive/overdense behavior.","section":"Section 2.2"}],"recommendation":"major_revision","confidential_remarks":"The paper is suitable for a data-release journal and I see no novelty or scope problem. The main concern is that the text-to-archive mapping contains several internal inconsistencies that are load-bearing for users; these are fixable with careful proofreading against the actual FlatHUB directory listing."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Know this before reading: this is a data release paper, and it's a good one. The new content is the archive itself: full snapshot histories (601 snapshots, spacing <~25 Myr) for the core FIRE-2 runs, DMO versions, MHD+ and cosmic-ray variants, a later-reionization set, four new massive halos, z=7 and z=9 high-redshift runs, and four DMO boxes for initial conditions. Most of the underlying simulations appeared in earlier papers, so the novelty is the public availability plus uniform catalogs and merger trees. That is a real contribution, and the paper makes the data easy to use with a clear directory table, per-simulation citation instructions, a license, and honest caveats about the spurious cosmic-ray heating bug in the base runs, missing snapshots, and simulations without AGN feedback.\n\nThe prose is careful and the internal accounting mostly holds up. My main concern is the text-to-archive mapping, which is the load-bearing part of any data release. The stress-test note is right about two small inconsistencies: Section 2.3 lists the z=9 suite as z9m11a,b,c,d,e plus z9m12a, while Table 1 lists z9m11a,b,c and z9m12a,b,c; and the abstract/section counts of SMC/LMC-mass and lower-mass galaxies don't cleanly reconcile with Table 1's mass bins. These are exactly the sorts of typos that can cost a user hours, and they should be fixed in revision. They don't undermine the release itself, since the archive can be checked directly, but they confirm the text is the fragile link.\n\nI can't verify from the manuscript that every stated file exists on FlatHUB in the described structure, but that is true of any data paper, and nothing in the text makes me suspect the archive is missing. The paper is transparent about known issues, which earns trust.\n\nWho is this for? Anyone using or planning to use FIRE-2 for galaxy formation work, especially at resolved ISM scales. It deserves a serious referee and a minor revision, not a desk reject. I'd cite it when I use the data.","headline":"A solid, honest data release that expands FIRE-2 to full snapshot histories plus DMO, MHD+, cosmic-ray, and later-reionization variants; the main issues are minor text-to-archive inconsistencies that should be fixed before publication.","tokens_in":17919,"tokens_out":2519,"would_cite":true,"duration_ms":28352,"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":"The FIRE-2 project's second data release makes 119 cosmological zoom-in galaxy simulations publicly available, with dense snapshots, physics-variant runs, halo catalogs, and merger trees.","keywords":["FIRE-2 simulations","galaxy formation","cosmological zoom-in","data release","public archive","halo catalogs","merger trees","MHD"],"falsifier":"A reader downloads the directory listing of the archive (e.g., the Core suite subdirectory) and checks that each Base Physics run has exactly 601 snapshots with spacings under ~25 Myr, that the MHD+ and cosmic-ray suites contain 16 and 14 simulations respectively, and that tree.hdf5 merger trees exist for the stated Core runs; any mismatch would falsify the release description.","tokens_in":16851,"feed_emoji":"🌌","tokens_out":8770,"duration_ms":84773,"temperature":0.7,"pith_summary":"This paper announces and documents the second public data release (DR2) of the FIRE-2 cosmological zoom-in simulations of galaxy formation. The authors assert that the public archive now contains 119 simulations across four suites, with most runs including every saved snapshot from z≈99 down to their final redshift, spaced no more than about 25 million years apart. The Core suite alone provides 601 snapshots for 14 Milky Way-mass, 5 SMC/LMC-mass, and 4 lower-mass galaxies, along with resimulations that alter single physical ingredients: dark-matter-only, later reionization, magnetic fields with conduction and viscosity, and cosmic-ray feedback. Most runs come with (sub)halo and galaxy catalogs at every snapshot, and most Core runs include full merger trees, so a galaxy's assembly can be traced continuously through time. If the archive matches this description, the release gives the community a uniform, high-cadence set of simulated galaxies for studying how baryons assemble and how feedback shapes them.","feed_headline":"FIRE-2 releases 119 galaxy simulations, all snapshots included","feed_subtitle":"Dense snapshots, physics variants, and merger trees let anyone trace galaxy assembly from cosmic dawn to today.","key_machinery":"The load-bearing machinery is the uniform snapshot cadence: releasing every saved output (spacing ≤ ~25 Myr) rather than every tenth output. That cadence, combined with the Rockstar (sub)halo catalogs at all snapshots and the Consistent Trees merger trees, turns the archive into a continuous record of each halo's assembly. The physics-variant resimulations (dark-matter-only, later reionization, MHD+, cosmic rays) are the controlled experiments: they re-run the same initial conditions with exactly one ingredient changed, so any difference in outcome is attributable to that ingredient.","core_discovery":"DR2 extends the initial FIRE-2 release (Wetzel et al. 2023) by publishing all currently stored snapshots for most simulations rather than a sparse subset. The central claim is that the archive now holds the complete time history of each run: snapshot time spacings are all below about 25 Myr from z≈99 to the final output. The Core suite reaches z=0 with 601 snapshots for each of 23 galaxies/halos; the Massive Halo suite reaches z=1 with 278 snapshots for 8 halos; the High Redshift suite includes 34 simulations to z=5, z=7, or z=9; and four dark-matter-only boxes provide initial-condition templates. In addition to the base physics, the release includes controlled resimulations that vary the ph","pith_inferences":["A natural extension the authors do not discuss: adding derived or reduced data products (e.g., precomputed density fields, star formation maps, or summary lightcones) would lower the barrier for observational astronomers who do not work with simulation snapshots.","The fine time sampling could serve as training data for machine-learning emulators or subgrid models, since star formation histories and merger rates are resolved at ~25 Myr resolution.","The single constant-diffusion cosmic-ray model is explicitly noted as not the preferred model; the release implicitly provides a benchmark against which the newer multi-energy-bin cosmic-ray FIRE runs can be compared.","If 25 Myr spacing is sufficient to capture bursty star formation, future simulations could save storage by adopting adaptive or coarser output cadence away from star-forming events; conversely, the archive tells us where the interesting times are."],"forward_implications":["For every Core-suite galaxy, users can follow the main progenitor branch from z=99 to z=0 without gaps, because the 601 snapshots and full merger trees are provided.","The physics-variant resimulations share initial conditions with the Base runs, so pairwise comparisons isolate the effect of reionization timing, magnetic fields, or cosmic-ray feedback.","The dense snapshot spacing (≲25 Myr) is comparable to the timescale of bursty star formation, so the data resolve individual star-formation and feedback episodes in the ISM.","The four dark-matter-only boxes, at mass resolutions differing by factors of two, allow users to generate new zoom-in initial conditions at arbitrary target resolution, extending the suite beyond the published halos.","All data are released under CC-BY, so downstream analyses and derived products can be published without restriction, subject to citation."],"supporting_citations":[{"why":"Defines DR1 and the FIRE-2 snapshot/catalog contents that DR2 extends; the paper directs users there for full model details.","marker":"Wetzel et al. (2023)"},{"why":"Defines the FIRE-2 physics model that all Base simulations use.","marker":"Hopkins et al. (2018)"},{"why":"Introduced the MHD+ and cosmic-ray models whose resimulations DR2 releases.","marker":"Hopkins et al. (2020)"},{"why":"Rockstar halo finder used to generate (sub)halo catalogs at all snapshots.","marker":"Behroozi et al. (2013a)"},{"why":"Consistent Trees algorithm used to build the full merger trees for most Core runs.","marker":"Behroozi et al. (2013b)"},{"why":"The ultraviolet background model that the Later Reionization runs modify.","marker":"Faucher-Giguère et al. (2009)"},{"why":"Introduced the m09 Later Reionization simulation that DR2 releases.","marker":"Gandhi et al. (2022)"},{"why":"Introduced the z=7 and z=9 high-redshift simulations now included in DR2.","marker":"Ma et al. (2019)"},{"why":"Introduced the FIRE-2 versions of the B1/B2/C1/C2 massive halos added to the Massive Halo suite.","marker":"Cochrane et al. (2023a)"}],"fun_headline_variants":["FIRE-2 DR2: 119 galaxies, every snapshot from z≈99 to present","Full time-resolved history of 119 galaxy simulations now public","FIRE-2: all 119 simulation timelines released, no gaps","Second FIRE-2 data drop: complete snapshot archive for 119 runs","From cosmic dawn to final output: FIRE-2 releases full time series"],"cache_read_input_tokens":2816,"weakest_assumption_plain":"The paper's central claim is about the contents of a public online archive, so the load-bearing premise is that the files actually exist there with the stated directory structure, snapshot counts, and notes.txt exceptions—something that cannot be verified from the manuscript text alone.","fun_headline_variants_meta":{"raw":{"variants":["FIRE-2 DR2: 119 galaxies, every snapshot from z≈99 to present","Full time-resolved history of 119 galaxy simulations now public","FIRE-2: all 119 simulation timelines released, no gaps","Second FIRE-2 data drop: complete snapshot archive for 119 runs","From cosmic dawn to final output: FIRE-2 releases full time series"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000752,"raw_usage":{"total_tokens":3241,"prompt_tokens":861,"completion_tokens":2380,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":605,"completion_tokens_details":{"reasoning_tokens":2279}},"tokens_in":605,"tokens_out":2380,"duration_ms":20672,"temperature":1.0,"reasoning_tokens":2279,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-05T22:38:39.158477+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"A reader downloads the directory listing of the archive (e.g., the Core suite subdirectory) and checks that each Base Physics run has exactly 601 snapshots with spacings under ~25 Myr, that the MHD+ and cosmic-ray suites contain 16 and 14 simulations respectively, and that tree.hdf5 merger trees exist for the stated Core runs; any mismatch would falsify the release description.","supporting_citations":[],"review_version":1}