REVIEW
An analytical model for non-local stochastic field-level galaxy bias on the sphere
Cosmology and Nongalactic Astrophysics · Submitted 8 Oct 2026 · 226 references
Pith's verdict: A cleanly derived, honestly reported bias model with a genuinely independent validation; the abstract's blanket O(10 Mpc) bispectrum claim needs narrowing to match the body's own numbers.
T0 · AI review
Why Pith says this
Highlight any text to comment on it.
Referee report
The paper
The paper claims
Under this model, galaxy bias is a random field on the sky: a power-spectrum-only calibration reproduces the three-point clustering of simulated galaxies down to about 10 Mpc comoving scales.
The reading
What carries the argument
The load-bearing object is the stochastic bias field in harmonic space, $b_{\ell m} = \beta_\ell + \gamma_{\ell m}$, with $\gamma_{\ell m}$ a complex Gaussian of variance $\sigma_\ell^2$ drawn independently of the matter coefficients: a diagonal, mode-by-mode linear map that is non-local through the scale dependence of $\beta_\ell$ and stochastic through $\gamma_{\ell m}$, while preserving homogeneity, realness, and isotropy in expectation. The identities that carry the argument are the closed-form estimators $\hat\beta_\ell = C^{\rm b}_\ell/C_\ell$ and $\hat\sigma^2_\ell = C^{bb}_\ell/C_\ell - \hat\beta^2_\ell$, which tie every parameter to two-point statistics alone, so the bispectrum and wavelet phase harmonics used for validation are genuinely independent checks. The constraint structure matters as much as the estimators: real-valued $\beta_\ell$, uncorrelated real and imaginary parts ($\rho = 0$, forced by isotropy) with equal variances for the $m>0$ modes, and the single real $m=0$ mode carrying the total variance. A supporting lemma shows that Poisson sampling of tracers cannot generate scale-dependent stochastic bias — a decomposition theorem for Poisson variables forces any such scatter to enter at the field-level mapping, which is exactly where BOLTS puts it. The wavelet phase harmonic statistics serve as the decisive out-of-sample probe because they measure amplitude-phase correlations and cross-scale mode coupling that two-point statistics cannot see.
Extended reading notes
Core claim
The core claim is that the simplest linear bias model compatible with the cosmological principle is necessarily both non-local and stochastic: a mode-by-mode map $\delta^{\rm b}_{\ell m} = b_{\ell m}\,\delta_{\ell m}$ with $b_{\ell m} = \beta_\ell + \gamma_{\ell m}$, where $\beta_\ell$ is real and scale-dependent and $\gamma_{\ell m}$ is a zero-mean complex Gaussian independent of the matter phases. Homogeneity forces $\langle b_{\ell m}\rangle$ to be real and independent of $m$; realness pairs the $m$ and $-m$ coefficients; and statistical isotropy is recovered in expectation, order by order — exactly at the two-point level and up to corrections of relative weight $1/(2\ell+1)$ in isotropically averaged third-order estimates, with the $m>0$ modes splitting the stochastic variance equally between real and imaginary parts and the single real $m=0$ mode carrying the full variance. The parameters admit closed-form estimators from power spectra alone, $\hat\beta_\ell = C^{\rm b}_\ell/C_\ell$ and $\hat\sigma^2_\ell = C^{bb}_\ell/C_\ell - \hat\beta^2_\ell$, which is what makes the three-point tests out-of-sample. Validated against the FLAMINGO 'L2p8_m9' hydrodynamical run, the model reproduces two-point statistics by construction, recovers the equilateral angular bispectrum within sample variance down to $\mathcal{O}(10\,\mathrm{Mpc})$ comoving scales for $M_{\rm h} > 10^{11.5}\,M_\odot$ over $0<z<3$, and tracks the wavelet phase harmonic moments in the low-mass, low-redshift regime where non-locality and stochasticity are strongest. The paper also states where the model runs out: in the high-mass, high-redshift bin the stochastic power is the discreteness of the tracers themselves, and the Gaussian term erases the amplitude-phase coherence that the deterministic point-source structure possesses, leaving 4–5$\sigma$ residuals in the phase-harmonic statistics.
Load-bearing premise
The load-bearing premise is that the bias acts as a diagonal, mode-by-mode multiplication in harmonic space with Gaussian fluctuations drawn independently of the matter field, so any genuine mode coupling from non-linear gravity — and any scatter that is actually coherent tracer discreteness rather than independent randomness — can only be mimicked stochastically, and the paper's own high-mass, high-redshift residuals mark where that mimicry runs out.
Editorial extensions
If this is right
- In the regime where tracer scatter is genuine, power-spectrum calibration alone is sufficient for field-level forward models that also get the three-point statistics right: the BOLTS equilateral bispectrum matches FLAMINGO within sample variance down to about 10 Mpc, and the analytic prediction of Equation (20) matches the mock ensemble at all masses and redshifts tested.
- Local deterministic bias prescriptions leave spatially coherent residuals that trace the filamentary cosmic web, while BOLTS residuals are consistent with pure stochastic scatter, so those standard prescriptions are wrong at the field level even where they match power spectra.
- Eight parameters per halo-mass bin — a generalised Lorentzian for $\beta_\ell$ and a sigmoid for $\sigma_\ell$, with two exponents fixed — interpolate smoothly across $0<z\leq3$, so galaxy fields can be generated at arbitrary redshifts in this range by interpolation alone, without re-calibrating the power spectra.
- Adjacent redshift shells are statistically independent by construction, so a tomographic extension must promote the per-shell stochastic variance to a cross-shell covariance; the paper identifies this as the step needed for three-by-two-point analyses.
- The same fitting functions provide analytic nuisance-parameter priors for galaxy clustering and galaxy-galaxy lensing likelihoods, separating deterministic bias, stochastic bias, and shot noise instead of folding them into one amplitude.
Reading between the lines
- A testable extension the paper does not pursue: because the stochastic term enters linearly and Gaussian, the model admits analytic marginalisation over the bias draws, which could yield a fast closed-form likelihood for tomographic galaxy clustering — worth checking against the sample-variance-dominated errors of this validation.
- The construction could in principle be chained, mapping matter to collapsed matter and then to galaxies, but the paper notes the first leg is dominated by non-linear collapse; a direct test would compare a chained two-step BOLTS against the single-step model in dark-matter-only runs where halo positions match the hydrodynamical ones.
- If the bispectrum-to-power-spectrum relation is as tightly fixed by $(\beta_\ell, \sigma_\ell)$ as Equation (20) suggests, measuring the equilateral bispectrum ratio in real survey data would estimate the stochastic bias amplitude without mock calibration — a consistency check wide-area surveys could perform.
- The heavier-than-Gaussian coefficient tails the paper reports at low mass and high multipole suggest the Gaussian draw, not the diagonal structure, is the next limiting assumption; replacing $\gamma_{\ell m}$ with a tailed distribution that preserves the fitted power spectra would isolate whether the phase-harmonic residuals come from the tails or from mode coupling.
Original abstract
We present the `Bias of Overdensities in Large-scale Tracers on the Sphere' (BOLTS) model, a novel analytic galaxy bias model for projected fields on the sphere that incorporates stochasticity and non-locality already at linear order while enforcing physicality and statistical isotropy. Rather than as a deterministic transfer function, we model galaxy bias as a statistically isotropic random field, allowing stochasticity to enter through fluctuations in the bias itself. The construction yields an explicit decomposition into a linear, non-local stochastic and deterministic components in harmonic space. We fit the model parameters to the projected power spectra measured in large-volume hydrodynamical simulations (FLAMINGO), enabling sampling of galaxy populations from simulated matter fields while conditioning on galaxy properties such as halo mass or redshift. By construction, the model reproduces the two-point statistics down to the simulation resolution, and without being calibrated on it, BOLTS recovers the equilateral bispectrum within the sample variance of the simulation down to comoving scales of $\mathcal{O}(10 \, \mathrm{Mpc})$ for halo masses $ M_{\rm{h}} > 10^{11.5}\,\rm{M}_{\odot} $ over redshifts $0 < z < 3$. Relative to local bias prescriptions, it more accurately captures amplitude-phase correlations and mode coupling in the measured wavelet phase harmonics wherever the stochasticity of the tracers exceeds their discreteness, including low redshifts and lower halo masses $M_{\rm{h}} = 10^{11.5}\,\rm{M}_{\odot}$ where non-linearities become important. This model enables efficient field-level forward modelling for simulation-based inference and model comparison with large datasets from forthcoming galaxy surveys including Euclid, Rubin LSST, and Roman, while improving the accuracy of galaxy bias modelling with respect to standard linear deterministic models.
The compiled record
The paper's own sentences, arranged, kept beside the review and separate from it. Each label is drawn from a closed vocabulary; the number beside it is the probability the classifier gave that label. Numbers in the tables are the authors', checked by code. No sentence in this record was written by a model.
The abstract, sentence by sentence
We present the `Bias of Overdensities in Large-scale Tracers on the Sphere' (BOLTS) model, a novel analytic galaxy bias model for projected fields on the sphere that incorporates stochasticity and non-locality already at linear order while enforcing physicality and statistical isotropy. Rather than as a deterministic transfer function, we model galaxy bias as a statistically isotropic random field, allowing stochasticity to enter through fluctuations in the bias itself. The construction yields an explicit decomposition into a linear, non-local stochastic and deterministic components in harmonic space. We fit the model parameters to the projected power spectra measured in large-volume hydrodynamical simulations (FLAMINGO), enabling sampling of galaxy populations from simulated matter fields while conditioning on galaxy properties such as halo mass or redshift. By construction, the model reproduces the two-point statistics down to the simulation resolution, and without being calibrated on it, BOLTS recovers the equilateral bispectrum within the sample variance of the simulation down to comoving scales of $\mathcal{O}(10 \, \mathrm{Mpc})$ for halo masses $ M_{\rm{h}} > 10^{11.5}\,\rm{M}_{\odot} $ over redshifts $0 < z < 3$. Relative to local bias prescriptions, it more accurately captures amplitude-phase correlations and mode coupling in the measured wavelet phase harmonics wherever the stochasticity of the tracers exceeds their discreteness, including low redshifts and lower halo masses $M_{\rm{h}} = 10^{11.5}\,\rm{M}_{\odot}$ where non-linearities become important. This model enables efficient field-level forward modelling for simulation-based inference and model comparison with large datasets from forthcoming galaxy surveys including Euclid, Rubin LSST, and Roman, while improving the accuracy of galaxy bias modelling with respect to standard linear deterministic models.
background contribution method result interpretation main claim
| # | sentence | role | the authors' own work | states the main conclusion | strength |
|---|---|---|---|---|---|
| s0 | We present the `Bias of Overdensities in Large-scale Tracers on the Sphere' (BOLTS) model, a novel analytic galaxy bias model for projected fields on the sphere that incorporates stochasticity and non-locality already at linear order while enforcing physicality and statistical isotropy. | contribution 0.99 | 0.95 | 0.24 | asserted 1.00 |
| s1 | Rather than as a deterministic transfer function, we model galaxy bias as a statistically isotropic random field, allowing stochasticity to enter through fluctuations in the bias itself. | method 0.91 | 0.90 | 0.07 | asserted 1.00 |
| s2 | The construction yields an explicit decomposition into a linear, non-local stochastic and deterministic components in harmonic space. | method 0.85 | 0.92 | 0.08 | asserted 1.00 |
| s3 | We fit the model parameters to the projected power spectra measured in large-volume hydrodynamical simulations (FLAMINGO), enabling sampling of galaxy populations from simulated matter fields while conditioning on galaxy properties such as halo mass or redshift. | method 0.95 | 0.96 | 0.10 | asserted 1.00 |
| s4 | By construction, the model reproduces the two-point statistics down to the simulation resolution, and without being calibrated on it, BOLTS recovers the equilateral bispectrum within the sample variance of the simulation down to comoving scales of $\mathcal{O}(10 \, \mathrm{Mpc})$ for halo masses $ M_{\rm{h}} > 10^{11.5}\,\rm{M}_{\odot} $ over redshifts $0 < z < 3$. | result 1.00 | 0.95 | 0.28 | asserted 1.00 |
| s5 | Relative to local bias prescriptions, it more accurately captures amplitude-phase correlations and mode coupling in the measured wavelet phase harmonics wherever the stochasticity of the tracers exceeds their discreteness, including low redshifts and lower halo masses $M_{\rm{h}} = 10^{11.5}\,\rm{M}_{\odot}$ where non-linearities become important. | result 0.94 | 0.91 | 0.19 | asserted 1.00 |
| s6 | This model enables efficient field-level forward modelling for simulation-based inference and model comparison with large datasets from forthcoming galaxy surveys including Euclid, Rubin LSST, and Roman, while improving the accuracy of galaxy bias modelling with respect to standard linear deterministic models. | interpretation 0.76 | 0.66 | 0.51 | asserted 0.96 |
What was examined, and what was not
- arXiv listing abstract v1
- full text
- figures and tables
- references and citing sentences
- relations to other papers
- P abstract
- S sections
- T body sentences
- C citations
- R cross-paper relations
Figures
Figures reproduced from arXiv:2610.12259 by the authors.
References
Discussion
Disagree with the review, or know something it missed? Leave a comment, highlight any passage above to comment on it, or challenge this review formally.