REVIEW 3 major objections 5 minor 132 references
Disentangling modified gravity and galaxy bias with field-level inference
T0 review · 3 major / 5 minor · reviewed 2026-07-12 · grok-4.5
Pith's one-line read Field-level analysis of the full galaxy map breaks the modified-gravity/bias degeneracy that power spectra cannot resolve.
desk verdict Clean controlled demo that field-level Poisson on n_g breaks the f(R)–β degeneracy under known phases; the idealisation is already their stated limit. read the letter →
The pith
A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.
The reading
What carries the argument
The full-field Poisson likelihood: each voxel is treated as an independent Poisson draw whose mean is the predicted galaxy count under a given (f_R0, β) model; the global log-likelihood is simply the sum over all voxels, thereby retaining both Fourier amplitudes and phases.
What would settle it
Repeat the identical mock analysis after freely sampling the initial phases (or after replacing the fixed-phase COLA realisations with a full Bayesian reconstruction that marginalises over them) and check whether the |f_R0|–β degeneracy reappears at the power-spectrum level of width.
Extended reading notes
Core claim
When initial phases are fixed and known, a Bayesian Poisson likelihood evaluated on the three-dimensional galaxy number-count field jointly constrains |f_R0| and the primary bias parameter β more tightly than a Gaussian power-spectrum likelihood and, crucially, breaks the strong degeneracy between the two parameters that is inherent to two-point statistics.
Load-bearing premise
The initial phases of the dark-matter density field are assumed known and fixed; the paper itself notes that marginalising over them on real data is expected to remove the pure-phase information that currently breaks the degeneracy.
Editorial extensions
If this is right
- Stage-IV surveys can extract substantially tighter joint constraints on gravity and bias by analysing the three-dimensional galaxy field rather than power spectra alone.
- Cosmic-web classifiers applied to the underlying density field can isolate which environments (especially voids and walls) drive gravity constraints.
- Phase information alone already prefers the correct |f_R0|; combining it with amplitudes yields the tightest joint posteriors.
- The same pipeline is modular and can incorporate field-level emulators, more complete bias models, weak lensing, and redshift-space distortions.
Reading between the lines
- Once initial phases must be marginalised, residual field-level gains will come mainly from non-Gaussian amplitude information rather than pure morphology, so hybrid likelihoods that keep higher-order moments may still outperform pure two-point analyses.
- The finding that n_g = 1 voxels in voids/walls are the most informative suggests that carefully selected void and wall catalogues could serve as cheaper, partially field-level probes even before full map-level inference is routine.
- Because the degeneracy-breaking is environment-dependent, any residual screening or assembly-bias mismodelling will appear first as spatially coherent residuals in under-dense regions, giving a practical diagnostic for model misspecification.
Signed reviews
Editorial analysis
A structured set of objections, weighed in public.
Referee Report
Summary. The paper develops a field-level Bayesian inference pipeline that evaluates a Poisson likelihood directly on the 3D galaxy number-counts field n_g, jointly constraining the Hu–Sawicki |f_R0| and the primary local bias parameter β. Dark-matter fields are generated with MG-extended COLA under fixed initial phases; galaxies are painted with a four-parameter non-linear bias model (only β free, N retuned to keep N_tot fixed). On controlled mocks the full-field analysis yields tighter, less degenerate posteriors than a Gaussian power-spectrum likelihood (Table 1, Fig. 2), isolates the contribution of Fourier phases (§5.4), attributes most of the gain to under-dense voxels via cosmic-web classification (§5.3), and demonstrates robustness to Poisson seeds, initial-condition realisations and number-count thresholding (§6).
Significance. If the controlled demonstration holds, the work supplies a concrete, reproducible illustration that non-Gaussian morphology and phase information can break the classic MG–bias degeneracy that limits two-point analyses. The explicit phase-only versus amplitude-only comparison, the voxel-level likelihood decomposition, and the suite of 100 Poisson + 5 IC robustness tests are genuine strengths that make the result falsifiable and useful for Stage-IV survey forecasts. The transparent scoping to fixed known phases (already flagged by the authors) keeps the claim honest while still charting a path toward BORG-style marginalisation and field-level emulators.
major comments (3)
- §5.4 and §7 correctly note that pure-phase information vanishes once initial phases are marginalised, yet the abstract and the final sentence of the conclusions still present the degeneracy-breaking result as a “powerful path forward for au next-generation surveys” without quantifying residual constraining power from non-Gaussian amplitudes alone. A short forecast (or at least a clear statement) of what survives after phase marginalisation is needed for the claim to remain load-bearing outside the idealised setting.
- §2.3 and §3.1 fix ρ and ε at the Jasche & Lavaux (2019) values and only vary β (with N retuned). Because the exponential void-suppression term is precisely the piece that couples most strongly to the under-dense voxels identified as the main drivers in §5.3, the reported degeneracy breaking may be optimistic relative to a fully free local (or non-local) bias model. At minimum the authors should show one additional run in which ρ and ε are also free, or justify why the present restriction does not affect the central comparison.
- The COLA implementation (§2.2) uses a modest number of time-steps and a linear-field screening approximation. Under-dense regions, which dominate the likelihood ratios in Tables 2–5, are exactly where chameleon screening is weakest and residual force errors are largest. A quantitative comparison of the FML-COLA density field against a full N-body MG run (or a published accuracy benchmark at the same resolution) for the void/wall voxels that drive the constraints would strengthen the claim that the reported |f_R0| posteriors are not resolution artefacts.
minor comments (5)
- Fig. 1 (right) correlation matrix and Fig. 3 power-spectrum ratios would benefit from explicit k-bin labels or a second x-axis in h Mpc^{-1} so that the scale dependence of the residual degeneracy is immediately readable.
- Eq. (3) and the surrounding text use both n_g and n_g interchangeably; a single consistent notation for the expected galaxy count field would improve readability.
- The Percival correction is cited (Percival et al. 2022) but the numerical value of the factor F applied to the 1024-realisation covariance is never stated; a one-line addition in §3.2.1 would aid reproducibility.
- Several sentences in the abstract and introduction still contain concatenated words (“Wepresentafield-level ldots”, “Non-linearstructureformation ldots”); these are residual typesetting artefacts that should be cleaned.
- Table 1 reports 95 % upper limits for the GR case and 68 % intervals for F6; a uniform confidence level (or an explicit note) would make the comparison cleaner.
Circularity Check
No significant circularity: controlled mock comparison of two likelihoods under fixed known phases is self-contained and does not reduce any claim to its inputs by construction.
full rationale
The paper's central demonstration is a side-by-side Bayesian comparison, on the same fixed-seed COLA mocks, of a Gaussian power-spectrum likelihood versus a voxel-wise Poisson field-level likelihood for the joint parameters (β, |f_R0|). Fiducial bias parameters are taken from the external literature (Neyrinck et al. 2014; Jasche & Lavaux 2019 row 11) and held fixed except for the free parameter β; N is retuned solely to keep N_tot constant so that likelihood differences arise only from spatial morphology. No quantity is fitted to a subset of the mock data and then re-labelled a prediction. The phase-only versus amplitude-only tests (Sec. 5.4) and the cosmic-web voxel attribution (Sec. 5.3) are diagnostic analyses of the same likelihood surface, not circular redefinitions. Self-citations (e.g., Hoyland et al. 2025 COLA weak-lensing, Saadeh et al. emulators) appear only in the outlook and are not load-bearing for the reported constraints. The fixed-phase idealisation is explicitly scoped as a limitation (abstract, Sec. 5.4, Sec. 7) rather than hidden. Consequently the derivation chain does not collapse to its inputs by construction, by self-citation, or by renaming.
Assumptions & free parameters
free parameters (4)
- β (primary galaxy bias)
- |f_R0| (Hu–Sawicki strength)
- N, ρ, ε (remaining bias parameters)
- Background cosmology (Ω_m, σ_8,
_s,
_s, h,
_s)
assumptions (4)
- domain assumption Galaxy counts in each voxel are independent Poisson draws given the mean predicted by the bias model.
- domain assumption COLA (FML-COLA) with the stated resolution and 2LPT initial conditions is sufficiently accurate for the scales analysed.
- domain assumption The Neyrinck et al. (2014) local-in-density bias model with three parameters fixed is an adequate description of the galaxy–matter relation for this proof-of-concept.
- ad hoc to paper Initial Gaussian phases are known and identical for all models and the mock data.
Cite this review
Pith. "Pith review of Disentangling modified gravity and galaxy bias with field-level inference." pith.science (2026). https://pith.science/paper/UGCNLM57
@misc{pith2026260703514,
author = {Pith},
title = {Pith review of: Disentangling modified gravity and galaxy bias with field-level inference},
year = {2026},
howpublished = {\url{https://pith.science/paper/UGCNLM57}},
note = {Machine review of arXiv:2607.03514}
}
abstract
We present a field-level inference framework for testing gravity with the large-scale structure that exploits the full information content of the galaxy distribution. Traditional analyses based on the power spectrum discard non-Gaussian and Fourier phase information, resulting in strong degeneracies between modified gravity (MG) and galaxy bias. Our approach overcomes this limitation by performing a Bayesian likelihood analysis directly on the three-dimensional galaxy number counts field, jointly constraining MG and bias parameters using both amplitudes and phases. As an illustrative application, we analyse mock data in the context of the $f(R)$ theory of gravity and a non-linear galaxy bias model. Non-linear structure formation is modelled using the COmoving Lagrangian Acceleration (COLA) method under different gravity strengths, parameterised by $f_{R0}$. The resulting dark matter fields are then mapped to mock galaxy catalogues via a non-linear bias prescription. We demonstrate that, with fixed and known initial phases, including non-Gaussian and phase information yields tighter constraints on both $f_{R0}$ and the primary bias parameter, $\beta$, relative to the power-spectrum-only analyses. Notably, the field-level approach breaks the degeneracies between MG and galaxy bias inherent to two-point statistics. Through a cosmic web classification into voids, walls, filaments and clusters, we find that under-dense regions are the primary drivers in distinguishing gravity models at the field level. Finally, we establish the robustness of our pipeline against variations in initial conditions, Poisson noise, and galaxy field thresholding, providing a powerful path forward for field-level tests of gravity with next-generation surveys.
Figures
Figures from the paper (6 more)
Reference graph
Works this paper leans on
-
[1]
Pylians: Python libraries for the analysis of numerical simulations
-
[2]
The Manticore Project II: Bayesian digital twins of cosmic structure across the SDSS and BOSS volumes. arXiv e-prints , keywords =. doi:10.48550/arXiv.2606.10020 , archivePrefix =. 2606.10020 , primaryClass =
-
[3]
Physical Bayesian modelling of the non-linear matter distribution: New insights into the nearby universe. , keywords =. doi:10.1051/0004-6361/201833710 , archivePrefix =. 1806.11117 , primaryClass =
-
[4]
Models of f(R) Cosmic Acceleration that Evade Solar-System Tests
Hu, Wayne and Sawicki, Ignacy. Models of f(R) Cosmic Acceleration that Evade Solar-System Tests. Phys. Rev. D. 2007. doi:10.1103/PhysRevD.76.064004. arXiv:0705.1158
-
[6]
Bye-bye, Local-in-matter-density Bias: The Statistics of the Halo Field Are Poorly Determined by the Local Mass Density. , keywords =. doi:10.3847/2041-8213/ad97b9 , archivePrefix =. 2405.00635 , primaryClass =
-
[7]
Testing Scale-Dependent Modified Gravity with DESI DR1
Testing Scale-Dependent Modified Gravity with DESI DR1. arXiv e-prints , keywords =. doi:10.48550/arXiv.2604.26915 , archivePrefix =. 2604.26915 , primaryClass =
-
[8]
DESI DR2 results. II. Measurements of baryon acoustic oscillations and cosmological constraints. , keywords =. doi:10.1103/tr6y-kpc6 , archivePrefix =. 2503.14738 , primaryClass =
-
[9]
Validation of the Scientific Program for the Dark Energy Spectroscopic Instrument. , keywords =. doi:10.3847/1538-3881/ad0b08 , archivePrefix =. 2306.06307 , primaryClass =
Show all 132 references
-
[10]
Overview of the Euclid mission
Euclid: I. Overview of the Euclid mission. , keywords =. doi:10.1051/0004-6361/202450810 , archivePrefix =. 2405.13491 , primaryClass =
- [11]
- [12]
- [13]
- [14]
-
[15]
Journal of High Energy Astrophysics , keywords =
Cosmology intertwined: A review of the particle physics, astrophysics, and cosmology associated with the cosmological tensions and anomalies. Journal of High Energy Astrophysics , keywords =. doi:10.1016/j.jheap.2022.04.002 , archivePrefix =. 2203.06142 , primaryClass =
- [16]
- [17]
- [18]
- [19]
- [20]
- [21]
-
[22]
, keywords =
The 2dF Galaxy Redshift Survey: spectra and redshifts. , keywords =. doi:10.1046/j.1365-8711.2001.04902.x , archivePrefix =. astro-ph/0106498 , primaryClass =
2001 doi
- [23]
- [24]
- [25]
- [26]
- [27]
- [28]
- [29]
- [30]
- [31]
- [32]
- [33]
- [34]
- [35]
- [36]
- [37]
- [38]
-
[39]
, keywords =
The cosmological simulation code GADGET-2. , keywords =. doi:10.1111/j.1365-2966.2005.09655.x , archivePrefix =. astro-ph/0505010 , primaryClass =
2005 doi
- [40]
- [41]
- [42]
- [43]
-
[44]
, keywords =
Bayesian calibrated significance levels applied to the spectral tilt and hemispherical asymmetry. , keywords =. doi:10.1111/j.1365-2966.2007.12707.x , archivePrefix =. 0706.3014 , primaryClass =
2007 doi
- [45]
- [46]
-
[47]
, keywords =
The Velocity Field Olympics: assessing velocity field reconstructions with direct distance tracers. , keywords =. doi:10.1093/mnras/staf1960 , archivePrefix =. 2502.00121 , primaryClass =
- [48]
- [49]
- [50]
- [51]
- [52]
-
[53]
, keywords =
Phase information and the evolution of cosmological density perturbations. , keywords =. doi:10.1046/j.1365-8711.2000.03086.x , archivePrefix =. astro-ph/9905250 , primaryClass =
2000 doi
- [54]
- [55]
-
[56]
Phase Transitions in the Early Universe: Theory and Observations , year = 2001, editor =
Large-scale Structure, Theory and Statistics. Phase Transitions in the Early Universe: Theory and Observations , year = 2001, editor =. doi:10.48550/arXiv.astro-ph/0103017 , archivePrefix =. astro-ph/0103017 , primaryClass =
-
[57]
, keywords =
The Statistics of Peaks of Gaussian Random Fields. , keywords =. doi:10.1086/164143 , adsurl =
-
[58]
, keywords =
Nonlinear Gravitational Evolution of Phases and Amplitudes in One-dimensional Cosmological Density Fields. , keywords =. doi:10.1086/171726 , adsurl =
-
[59]
, keywords =
Information content of the non-linear matter power spectrum. , keywords =. doi:10.1111/j.1745-3933.2005.00051.x , archivePrefix =. astro-ph/0502081 , primaryClass =
2005 doi
- [60]
- [61]
- [62]
- [63]
- [64]
- [65]
- [66]
- [67]
- [68]
-
[69]
Bayesian Large Scale Structure inference and cosmic web analysis
- [70]
- [71]
- [72]
- [73]
- [74]
- [75]
- [76]
-
[77]
The Zeldovich Universe: Genesis and Growth of the Cosmic Web , year = 2016, editor =
Distinguishing f(R) gravity with cosmic voids. The Zeldovich Universe: Genesis and Growth of the Cosmic Web , year = 2016, editor =. doi:10.1017/S1743921316010632 , archivePrefix =. 1410.0133 , primaryClass =
- [78]
- [79]
- [80]
- [81]
- [82]
- [83]
- [84]
-
[85]
, keywords =
Lagrangian perturbation theory - A key-model for large-scale structure. , keywords =
-
[86]
, year = 1970, month = mar, volume =
Gravitational instability: An approximate theory for large density perturbations. , year = 1970, month = mar, volume =
1970
-
[87]
Theory and implementation
The persistent cosmic web and its filamentary structure - I. Theory and implementation. , keywords =. doi:10.1111/j.1365-2966.2011.18394.x , archivePrefix =. 1009.4015 , primaryClass =
2011 doi
- [88]
-
[89]
, keywords =
On the spatial correlations of Abell clusters. , keywords =. doi:10.1086/184341 , adsurl =
- [90]
- [91]
- [92]
- [93]
- [94]
- [95]
- [96]
- [97]
- [98]
- [99]
- [100]
- [101]
-
[102]
arXiv e-prints , keywords =
Cosmology inference with perturbative forward modeling at the field level: a comparison with joint power spectrum and bispectrum analyses. arXiv e-prints , keywords =. doi:10.48550/arXiv.2509.09673 , archivePrefix =. 2509.09673 , primaryClass =
-
[103]
Simulations and nonlinearities beyond CDM
Euclid preparation: LXXI. Simulations and nonlinearities beyond CDM. 3. Constraints on f(R) models from the photometric primary probes. , keywords =. doi:10.1051/0004-6361/202452184 , archivePrefix =. 2409.03524 , primaryClass =
- [104]
- [105]
- [106]
- [107]
- [108]
- [109]
-
[110]
, keywords =
The 2M++ galaxy redshift catalogue. , keywords =. doi:10.1111/j.1365-2966.2011.19233.x , archivePrefix =. 1105.6107 , primaryClass =
2011 doi
- [111]
- [112]
- [113]
- [114]
- [115]
-
[116]
Testing gravity with the full-shape galaxy power spectrum: First constraints on scale-dependent modified gravity
Aviles, Alejandro. Testing gravity with the full-shape galaxy power spectrum: First constraints on scale-dependent modified gravity. Phys. Rev. D. 2025. doi:10.1103/PhysRevD.111.L021301. arXiv:2409.15640
-
[117]
Testing f(R) Gravity from Cosmic Shear Measurements
Bai, Jiachen and Xia, Jun-Qing and Zhao, Gong-Bo. Testing f(R) Gravity from Cosmic Shear Measurements. Astrophys. J. 2025. doi:10.3847/1538-4357/ae07c4. arXiv:2412.04807
2025 doi
- [118]
-
[119]
and Novaes, Camila P
Armijo, Joaquin and Marques, Gabriela A. and Novaes, Camila P. and Thiele, Leander and Cowell, Jessica A. and Grand \'o n, Daniela and Shirasaki, Masato and Liu, Jia. Cosmological constraints using Minkowski functionals from the first year data of the Hyper Suprime-Cam. Mon. N...
- [120]
- [121]
-
[122]
and Pontzen, Andrew and Jasche, Jens and Lavaux, Guilhem
Stopyra, Stephen and Peiris, Hiranya V. and Pontzen, Andrew and Jasche, Jens and Lavaux, Guilhem. Towards accurate field-level inference of massive cosmic structures. Mon. Not. Roy. Astron. Soc. 2023. doi:10.1093/mnras/stad3170. arXiv:2304.09193
- [123]
-
[124]
A field-level reaction for screened modified gravity
Saadeh, Daniela and Koyama, Kazuya and Morice-Atkinson, Xan. A field-level reaction for screened modified gravity. Mon. Not. Roy. Astron. Soc. 2025. doi:10.1093/mnras/staf1605. arXiv:2506.18876
2025 doi
-
[125]
Field-level emulation of cosmic structure formation with cosmology and redshift dependence
Jamieson, Drew and Li, Yin and Villaescusa-Navarro, Francisco and Ho, Shirley and Spergel, David N. Field-level emulation of cosmic structure formation with cosmology and redshift dependence. JCAP. 2025. doi:10.1088/1475-7516/2025/03/072. arXiv:2408.07699
-
[126]
, keywords =
On the connection between field-level inference and n-point correlation functions. , keywords =. doi:10.1088/1475-7516/2025/09/056 , archivePrefix =. 2504.15351 , primaryClass =
2025 doi
- [127]
- [128]
- [129]
- [130]
- [131]
- [132]
- [133]
Reviewed July 12, 2026 · model on record in the stance chip above.
Discussion (0). Continue with ORCID to comment.