FlowingFlowing
← Back to all development cases
This case was judged by an anonymized AI reviewer and has not been independently reviewed by people; inspect the draft, cards, and both outputs to assess it.
AutocompleteWith cards better200 / 280 · beu-714d344fba0dd9bb

The halo model for cosmology: a pedagogical review

Astrophysics · 2303.08752v2

FLOWING EVIDENCE BENCHMARK

How do we compare writing with and without cards?

At the same writing position, the same model produces one first result with cards and one without. All 280 pairs enter the overall anonymized review, which assesses whether a decision not to continue fits the draft.

Overall: 138 with-card wins, 26 ties, 107 no-card wins, 5 pairs where neither was usable, and 4 inconclusive judgments. Both sides wrote text in 218 pairs; 48 had a deliberate stop and 14 included an invalid output.

SAME DRAFT · WITH OR WITHOUT CARDS

01 · FIXED WRITING POSITION

MANUSCRIPT

Same writing position ▌

Matched autocomplete at the same draft position

02 · TWO MATCHED VERSIONS

SHARED BY BOTH

Draft, writing task, and generation model

A · With cards

Relevant literature cards supplied

B · Without cards

No literature cards supplied

Generation model

Same model and version

A → first output

B → first output

Anonymized AI judge

First outputs are anonymized as X and Y

Continuation review: accuracy, fit to the writing task and usability
Returns: X preferred / tie / Y preferred / both unusable

Review again with the answer order swapped

HOW DOES BLIND JUDGING WORK?

① Anonymize both outputs
The AI judge sees the same draft and both first outputs without knowing which received cards.

② Compare and swap order
The judge applies task-specific criteria in X/Y and then Y/X order.

③ Review disagreements
A third pass resolves disagreements. Incomplete reviews remain in the denominator.

How are positions and cards selected?

We selected 40 writing positions in each of seven fields, for 280 comparisons. The position and relevant literature cards were set before each comparison, so both versions began from the same draft.

The main result retains every frozen position, including places where cards might not add a useful next move. This avoids selecting cases based on their outputs, but 49% describes the whole cohort, not a win rate where evidence clearly had an opportunity to help. This cohort did not record that opportunity label for each position before generation, so those two groups cannot yet be separated reliably.

CARD USE · 140/280

Reported card use in continuations with text

The model reports whether a card affected its continuation. In 140 of 280 tests it produced continuation text and reported card use. The previous count of 150 included 10 cases without continuation text and has been corrected.

COMPARISON · 70 / 7 / 30

Which written version was better?

Of the 107 reported-use cases where both versions produced text, the blinded AI judge preferred the with-card version 70 times, found 7 about the same, and preferred the without-card version 30 times.

How are the two table percentages calculated?

ADOPTION COVERAGE

140reported use with text
÷
280all positions
=
50%

WITH-CARD VERSION PREFERRED

70with cards better
÷
107comparable tests
=
65%

The percentages use different groups: 50% counts continuations with text reporting card use among all 280 tests; 65% covers the 107 such cases with text on both sides. The generating model reports card use, while an anonymized AI judge compares writing quality.

Card use is self-reported by the generator. The anonymized AI judge sees the draft and source snippets, so source-aware preference remains possible; there was no independent human review. The overall result describes these 280 frozen positions, not a win probability for any writing task.

Manuscript writing position

Text excerpt · not a PDF page

Research manuscript · excerpt

The halo model for cosmology: a pedagogical review

1 Introduction

…lapse in exotic cosmologies, the halo model also provides predictions for non-linear clustering in beyond- Λ \Lambda CDM models. The halo model has been widely used to model observations of a variety of large-scale structure probes, most notably as the primary technique to model the underlying non-linear matter power spectrum. By documenting these varied and often distinct use cases, we seek to further coherent halo model analyses of future multi-tracer observables. This review is accompanied by the release of pyhalomodel , flexible software to conduct a wide range of halo-model calculations.

…oes, which are 𝒪 ⁡ ( 100 ) \mathcal{O}(100) times denser than the cosmological average – a view that has been largely corroborated by numerical simulations. Once the properties and distribution of these haloes are known, one can estimate the statistical properties of the matter distribution in the cosmos. To be concrete, power spectra for matter and its tracers can be understood as the sum of two components: inter-halo (two-halo) and intra-halo (one-halo) clustering. The halo model, therefore, can (and has) been used in analysing cosmological data from various probes of large-scale structure.

↓ Next paragraph was generated here

The highlighted text is near the writing position. This is a layout of the evaluation input, not a PDF screenshot.

A matched comparison at the same writing positionOverall AI review: With cards better

Both arms use the same model, task and manuscript context. Only the retrieved passages differ. The first recorded output from each arm is shown below.

Read overall AI review rationale

A = without cards · B = with cards

The draft ends with a paragraph introducing the basic halo model framework (haloes as dense regions, two-halo vs one-halo terms, applications to cosmological probes). The preceding paragraph mentions documenting varied use cases and releasing pyhalomodel. Output A abstains with 'complete', but the draft clearly needs a next move: the basic framework has just been introduced, and a review paper would typically continue by elaborating on extensions, refinements, or the scope of what the review covers. The blank paragraph position is an active continuation point, not a natural ending. Output B provides a coherent continuation that naturally follows from the introduction: it lists important extensions of the halo model (substructure, concentration scatter, smooth mass component) that are well-supported by the sources and relevant to the review's stated goal of documenting 'varied and often distinct use cases.' This bridges toward more detailed discussion and fits the academic register. While B is somewhat brief, it makes a concrete, accurate next move that could be inserted without repair. A's abstention is premature because the manuscript has only just set up the basic framework and needs to develop it further before concluding or transitioning.

Evidence

Same model and position · retrieved passages supplied · first output

Preferred by blind review

The model has been extended in several directions, including the treatment of substructure within haloes, the scatter of halo concentrations at fixed mass, and the smooth component of uncollapsed mass.

Passages supplied to the Evidence version

Card 1 · Testing the Warm Dark Matter paradigm with large-scale structures

In § III we developed the halo model of large-scale structure. We extended the model by including two components of clustered mass in the Universe. That which is in collapsed dark matter haloes and that which is in the form of smooth uncollapsed mass. This extended model requires that we understand…

Read full passage excerpt

In § III we developed the halo model of large-scale structure. We extended the model by including two components of clustered mass in the Universe. That which is in collapsed dark matter haloes and that which is in the form of smooth uncollapsed mass. This extended model requires that we understand three correlation functions: the halo-halo, smooth-halo and smooth-smooth correlation functions. The halo-halo clustering can be calculated as in the standard halo model.

Card 2 · Halo model description of the non-linear dark matter power spectrum at k≫ 1 Mpc−1

Accurate knowledge of the non-linear dark-matter power spectrum is important for understanding the large-scale structure of the Universe, the statistics of dark-matter haloes and their evolution, and cosmological gravitational lensing. We analytically model the dark-matter power spectrum and its cro…

Read full passage excerpt

Accurate knowledge of the non-linear dark-matter power spectrum is important for understanding the large-scale structure of the Universe, the statistics of dark-matter haloes and their evolution, and cosmological gravitational lensing. We analytically model the dark-matter power spectrum and its cross-power spectrum with dark-matter haloes. Our model extends the halo-model formalism, including realistic substructure population within individual dark-matter haloes and the scatter of the concentration parameter at fixed halo mass. We consider three prescriptions for the mass-concentration relation and two for the substructure distribution in dark-matter haloes. We show that this extension of the halo model mainly increases the predicted power on the small scales, and is crucial for proper modeling the cosmological weak-lensing signal due to low-mass haloes. Our extended formalism shows how the halo model approach can be improved in accuracy as one increases the number of ingredients that …

Card 3 · Halo model and halo properties in Galileon gravity cosmologies

We investigate the performance of semi-analytical modelling of large-scale structure in Galileon gravity cosmologies using results from N-body simulations. We focus on the Cubic and Quartic Galileon models that provide a reasonable fit to CMB, SNIa and BAO data. We demonstrate that the Sheth-Tormen…

Read full passage excerpt

We investigate the performance of semi-analytical modelling of large-scale structure in Galileon gravity cosmologies using results from N-body simulations. We focus on the Cubic and Quartic Galileon models that provide a reasonable fit to CMB, SNIa and BAO data. We demonstrate that the Sheth-Tormen mass function and linear halo bias can be calibrated to provide a very good fit to our simulation results. We also find that the halo concentration-mass relation is well fitted by a power law. The nonlinear matter power spectrum computed in the halo model approach is found to be inaccurate in the mildly nonlinear regime, but captures reasonably well the effects of the Vainshtein screening mechanism on small scales. In the Cubic model, the screening mechanism hides essentially all of the effects of the fifth force inside haloes. In the case of the Quartic model, the screening mechanism leaves behind residual modifications to gravity, which make the effective gravitational strength time-varyin …

Card 4 · Nonlinear power spectra of dark and luminous matter in the halo model of structure formation

The late stages of large-scale structure evolution are treated semianalytically within the framework of the modified halo model. We suggest a simple yet accurate approximation for relating the nonlinear amplitude to the linear one for spherical density perturbation. For halo concentration parameter…

Read full passage excerpt

The late stages of large-scale structure evolution are treated semianalytically within the framework of the modified halo model. We suggest a simple yet accurate approximation for relating the nonlinear amplitude to the linear one for spherical density perturbation. For halo concentration parameter $c$, a new computation technique is proposed, which eliminates the need for interim evaluation of the ${z}_{\mathrm{col}}$. The validity of the technique is proved for $\ensuremath{\Lambda}\mathrm{CDM}$ and $\ensuremath{\Lambda}\mathrm{WDM}$ cosmologies. Also, the parameters for the Sheth-Tormen mass function are estimated. The modified and extended halo model is applied for determination of the nonlinear power spectrum of dark matter, as well as for a galaxy power spectrum estimation. The semianalytical techniques for the dark matter power spectrum are verified by comparison with data from numerical simulations. Also, the predictions for the galaxy power spectra are confronted with ''observ …

Card 5 · Substructure and the halo model of large-scale structure

We develop the formalism to include substructure in the halo model of clustering. Real haloes are not likely to be perfectly smooth, but have substructure which has, to date, been neglected in the halo model – our formalism allows one to estimate the effects of this substructure on measures of clust…

Read full passage excerpt

We develop the formalism to include substructure in the halo model of clustering. Real haloes are not likely to be perfectly smooth, but have substructure which has, to date, been neglected in the halo model – our formalism allows one to estimate the effects of this substructure on measures of clustering. We derive expressions for the two-point correlation function, the power-spectrum, the cross-correlation between galaxies and mass, as well as higher order clustering measures. Simple forms of the formulae are obtained for the limit in which the size of the substructure and the mass fraction in it is small. Inclusion of substructure allows for a more accurate analysis of the statistical effects of gravitational lensing. It may also bring the halo model predictions into better agreement with the small-scale structure seen in recent high-resolution simulations of hierarchical clustering.

Model-reported card use

Card 5 · fact

Affected output: the treatment of substructure within haloes

Model explanation: This row supplied the concrete fact that substructure has been incorporated as an extension of the halo model.

Card 2 · fact

Affected output: the scatter of halo concentrations at fixed mass

Model explanation: This row supplied the specific extension of including scatter in the concentration parameter at fixed halo mass.

Card 1 · fact

Affected output: the smooth component of uncollapsed mass

Model explanation: This row supplied the fact that the halo model has been extended to include a smooth uncollapsed mass component.

No Evidence

Same model and position · no retrieved passages · first output

Not preferred in this comparison

No continuation text was produced.

Model action：complete

Earlier card-use comparison

The earlier diagnostic compared only cases with reported card use and text on both sides; its scope differs from the overall AI review above.

The with-card arm produced text and reported card use, but the no-card arm produced no text. This counts as reported use, without entering the two-text preference comparison.

The passages are input material; their presence does not mean an output used them or that they support every claim. Draft and source passages are excerpted; no simulated scores or PDF appear here.