---
title: 'GRAS: Polysemous Terms in Research'
url: https://www.emergentmind.com/topics/gras
type: topic
---

# GRAS: Polysemous Terms in Research

GRAS denotes several unrelated constructs in contemporary research literature. In algebraic number theory, it appears in the **Gras conjecture**, which relates unit groups and ideal class groups in abelian extensions [1102.0903]. In high-dimensional statistics, the closely related acronym **GRASS** denotes **graphical sure screening** for Gaussian graphical models [1407.7819]. In astrophysics, **GRAS** denotes **GRavitational Anti-Screening**, a dark-matter alternative formulated through vacuum-induced dipole effects [2602.09249]. The string also appears in **Multi-GraS**, a multiplex graph model for extractive summarization [2108.12870], and in the **GRAS** benchmark for demographic bias in vision-language models, where it expands to **Gender, Race, Age, and Skin Tone** [2508.18989]. The term is therefore intrinsically polysemous, and its meaning is fixed by disciplinary context.

## 1. Gras conjecture in algebraic number theory

In its classical number-theoretic usage, GRAS refers to the **Gras conjecture**, which predicts a precise relationship between the unit group and ideal class group of an abelian extension, measured using the so-called “index.” For a rational character $\chi$ of $G=\mathrm{Gal}(K/k)$, the conjecture is summarized in the form
$$
[\mathbb{Z}[G]e_\chi:\ell_K(E_K)e_\chi]=|\mathrm{Cl}(K)[e_\chi]|.
$$
The central objects are the ideal class group $\mathrm{Cl}(K)$), the unit group $\mathcal{O}_K^\times$, and distinguished unit subgroups generated from elliptic, Stark, or circular units, depending on the arithmetic setting [1102.0903].

For **abelian extensions of an imaginary quadratic field** $k$, “On Gras conjecture for imaginary quadratic fields” proves the previously unresolved case where the prime $p$ divides the number of roots of unity in $k$ [1102.0903]. The paper extends Rubin’s methods, defines $E_K$ as the subgroup of $K^\times$ generated by the roots of unity in $k$ and all possible norms of Stark units, and combines Euler-system constructions with elementary group-theoretic and index calculations via the generalized index of Sinnott. Its main theorem states that if $p$ divides $w_k$ and does not divide $[K:k]$, and if $x$ is a nontrivial irreducible $\mathbb{Z}_p$-character of $G$, then
$$
[e_x(\mathbb{Z}_p\otimes \mathcal{O}_K^\times):e_x(\mathbb{Z}_p\otimes E_K)]
=
|e_x(\mathbb{Z}_p\otimes \mathrm{Cl}(K))|.
$$
This completes the proof of the Gras conjecture for all primes in that setting [1102.0903].

A function-field analogue is established in “The Gras conjecture in function fields by Euler systems” [1102.0617]. There the ambient objects are a global function field $k$, a finite abelian extension $K\subset k_\infty$, the unit group $O_K^\times$, the class group $\mathrm{Cl}(O_K)$, and a Stark-unit subgroup $E_K$. The proof uses Euler systems constructed from torsion points of sign-normalized Drinfeld modules and adapts the techniques of Thaine, Kolyvagin, and Rubin. Under the paper’s explicit hypotheses on $p$ and $\chi$, the main equality is
$$
[e_\chi(\mathbb{Z}_p\otimes O_K^\times):e_\chi(\mathbb{Z}_p\otimes E_K)]
=
\# e_\chi(\mathbb{Z}_p\otimes \mathrm{Cl}(O_K)).
$$
This places Stark units in global function fields in the same structural role that elliptic and cyclotomic units occupy in number fields [1102.0617].

A further extension appears in “Gauss Sums, Stickelberger’s Theorem, and the Gras Conjecture for Ray Class Groups” [1502.01578]. For a real abelian number field $k$, an odd prime $p$ not dividing $[k:\mathbb{Q}]$, the unit group
$$
E_d=\{\varepsilon\in \mathcal{O}_k^\times:\varepsilon\equiv 1 \pmod d\},
$$
the subgroup $C_d$ of $d$-circular units, and the ray class group $\mathfrak{C}(d)$, the paper proves a ray-class version of the conjecture:
$$
|e_\rho \operatorname{Syl}_p(E_d/C_d)|
=
|e_\rho \operatorname{Syl}_p(\mathfrak{C}(d))|
$$
when the ramification index of $p$ in $k$ is less than $p-1$ [1502.01578]. It also constructs explicit Galois annihilators of $\mathrm{Syl}_p(\mathfrak{C}_{\mathfrak{a}})$ akin to the classical Stickelberger theorem. Across these variants, the common theme is the comparison of explicit unit subgroups with class-theoretic invariants via characterwise index formulas.

## 2. GRASS in Gaussian graphical models

In high-dimensional statistics, the near-homographic term **GRASS** stands for **graphical sure screening** [1407.7819]. It is a very simple and computationally efficient screening procedure for recovering the structure of a Gaussian graphical model in the high-dimensional setting. Rather than estimating the precision matrix directly, GRASS thresholds the elements of the sample covariance or correlation matrix. With standardized variables and data matrix $\mathbf{X}$, the empirical correlation for features $a,b$ is
$$
r_{ab}=\frac{\mathbf{X}_a^T\mathbf{X}_b}{n},
$$
and the estimated edge set is
$$
\widehat{\mathcal{E}}_{\gamma_n}
=
\{(a,b):a<b,\ |r_{ab}|>\gamma_n\}.
$$
An edge is declared between $a$ and $b$ if and only if $|r_{ab}|>\gamma_n$ [1407.7819].

Its principal theoretical result is the **sure screening property**: with very high probability, the GRASS estimated edge set contains the true edge set. Under the condition that, for a true edge $(a,b)$, the population covariance satisfies $|\sigma_{ab}|\ge C_1 n^{-\kappa}$ with $\kappa\in(0,1/2)$, and when $\log p=C_3 n^\xi$ with $\xi\in(0,1-2\kappa)$, choosing
$$
\gamma_n=\frac{2}{3}C_1 n^{-\kappa}
$$
yields
$$
\mathbb{P}\big(\mathcal{E}\subseteq \widehat{\mathcal{E}}_{\gamma_n}\big)
\ge
1-C_4\exp(-C_5 n^{1-2\kappa}).
$$
The paper also shows that, under an additional eigenvalue-growth condition, the size of the estimated neighborhood is controlled, which makes the screening step computationally useful in ultra-high dimensions [1407.7819].

The method includes an analytical threshold for expected false positive rate control. If $f$ is the maximum number of acceptable false positives, one sets
$$
\gamma_n=
\frac{\Phi^{-1}\!\left(1-\frac{f}{p(p-1)}\right)}{\sqrt{n}},
$$
and under mild assumptions the expected false positive rate satisfies
$$
\mathbb{E}[\mathrm{FPR}] \le \frac{f}{|\mathcal{E}^c|}.
$$
The computational cost is $O(p^2)$, in contrast with the $O(p^3)$ cost stated for the graphical lasso and neighborhood selection. The paper further notes that the connected components produced by thresholding the sample covariance at level $\gamma_n$ correspond exactly to those recovered by the graphical lasso with $\lambda=\gamma_n$ [1407.7819]. Simulation studies and a gene-expression example show that GRASS performs competitively with more complex methods, especially when the sparsity pattern of the correlation matrix mirrors that of the precision matrix [1407.7819].

## 3. GRAS as gravitational anti-screening

In astrophysics, **GRAS** denotes **GRavitational Anti-Screening**, a theory presented as an alternative to dark matter and to MOND’s AQUAL formulation [2602.09249]. Its core physical picture is that the gravitational field of baryonic matter induces mass dipole moments in quantum vacuum fluctuations, producing an anti-screening effect that boosts the effective observed gravitational mass. The field equation is written as
$$
\nabla^2\Phi = 4\pi G(\rho_{\text{bar}}+\rho_{\text{dipole}}),
$$
with dipole density
$$
P = \frac{1}{4\pi G}f\!\left(\frac{g}{g_0}\right)g,
$$
and effective dipole contribution
$$
\rho_{\text{dipole}}=-\nabla\cdot P.
$$
Under spherical symmetry this reduces to
$$
g_G = g_{\text{bar}} + f\!\left(\frac{g_G}{g_0}\right)g_G,
$$
or equivalently
$$
(1-f)g_G=g_{\text{bar}}.
$$
A specific interpolating function cited as well matched to the galactic radial acceleration relationship is
$$
f(x)=(1+x^{-2})^{-1}.
$$
The paper explicitly states that GRAS and AQUAL are mathematically identical after identifying $\mu(x)=1-f(x)$ [2602.09249].

The same manuscript argues that both AQUAL and standard GRAS fail to explain the bulk of the missing mass in galaxy clusters, even though they are in excellent agreement with observations of galaxies, including the galactic RAR [2602.09249]. The proposed modification introduces sensitivity to the radial variation of the field through
$$
\beta(r)\equiv -\frac{d\ln g}{d\ln r}
=
-\frac{r}{g}\frac{dg}{dr},
$$
and replaces the standard spherical equation by
$$
g_{GA}=g_{\text{bar}}+\beta\, f\!\left(\frac{g_{GA}}{g_0}\right)g_{GA}.
$$
The corresponding dipole interpretation becomes
$$
P=-\frac{1}{4\pi G}\beta\, f\!\left(\frac{g_{GA}}{g_0}\right)g_{GA}.
$$
The modified field is solved with an outer boundary condition chosen to match the galactic RAR regime [2602.09249].

According to the paper, the modified equation has **just a single free parameter**, the universal acceleration scale $g_0$, and is applied to a sample of six relatively bound galaxy clusters [2602.09249]. It is reported to fit the observationally inferred dynamical mass profiles far better than the unmodified equation, while retaining galactic behavior and satisfying solar-system constraints. A central interpretation is that the full baryonic density profile, rather than only the enclosed baryonic mass, becomes decisive for the gravitational response. This suggests a route for explaining the difference between the galactic RAR and the cluster RAR within a single modified-field framework [2602.09249].

## 4. GRAS-related graph models in machine learning and computational biology

The string “GraS” also appears in **Multi-GraS**, a graph-based model for extractive text summarization [2108.12870]. Multi-GraS is built on a **Multiplex Graph Convolutional Network (Multi-GCN)** that jointly models multiple types of relationships among sentences and words. At the word level, it uses intra-sentential syntactic and semantic graphs; at the sentence level, it uses inter-sentential natural-connection and semantic-similarity graphs. The sentence-level natural connection is defined through shared keywords,
$$
\mathbf{A}_{nat}[m,m']
=
\sum_{w\in\mathcal{W}}
\mathrm{tfidf}_{(s_m,w)}\cdot \mathrm{tfidf}_{(s_{m'},w)},
$$
while semantic sentence similarity is
$$
\mathbf{A}_{sem_s}[m,m']
=
|\mathbf{x}_{s_m}^T\mathbf{x}_{s_{m'}}|.
$$
For each relation $r$, Multi-GCN applies a separate GCN with an inner skip connection,
$$
\hat{\mathbf{H}}_r^{(l)}
=
\mathrm{GCN}_r^{(l)}(\mathbf{A}_r,\mathbf{H}_r^{(l-1)})
+
\mathbf{H}_r^{(l-1)},
$$
followed by relation aggregation and an outer skip connection [2108.12870].

The model is evaluated on the CNN/DailyMail benchmark. The reported ROUGE scores are **43.16** for R-1, **20.14** for R-2, and **39.49** for R-L, compared with **42.95**, **19.76**, and **39.23** for HSG in the reported table [2108.12870]. The paper states that removing the multiplex GCN from either the word or sentence block reduces performance, that both inner and outer skip connections are crucial, and that all modeled relation types contribute to the final result. In that literature, “GraS” is therefore a model name tied to graph-based summarization rather than to arithmetic or gravity [2108.12870].

A related acronymic development is **GRASMOS**, short for **Graph Signage Model Selection for Gene Regulatory Networks** [2211.09642]. GRASMOS addresses signed directed networks by fitting sign-generation models to a fixed graph topology through maximum likelihood. The latent structure assigns each node a group label $C(v)$, typically in $S=\{\text{Activator},\text{Repressor}\}$, and the likelihood is
$$
\mathcal{L}(\Theta)=P(A\mid \Theta)
=
\sum_C
\left[
\prod_{(u,v)\in E}
P_{C(u),C(v)}^{A(u,v)}\cdot P(C)
\right].
$$
The framework defines Node-Oblivious, Source-Consistent, Target-Consistent, and Bi-Node-Consistent models, with algorithms ranging from closed-form expressions to MCMC sampling. The paper reports evaluation on synthetic datasets and on real-world GRNs including *E. coli* and *Bacillus subtilis*, with BNC models achieving the best negative log-likelihood on the cited real datasets [2211.09642]. Although GRASMOS is not itself “GRAS,” it illustrates how the same letter sequence recurs in graph-centric modeling nomenclature.

## 5. GRAS as a benchmark for demographic bias in vision-language models

In multimodal ML, **GRAS** expands to **Gender, Race, Age, and Skin Tone** and denotes a benchmark for measuring demographic bias in vision-language models [2508.18989]. The benchmark is designed for visual question answering and is presented as offering the most diverse coverage to date across four demographic attributes. Its dataset consists of **5,010 stratified face images** from FairFace and AI-Face, **100 personality trait words**—50 positive and 50 negative—and **five linguistically diverse, yet semantically equivalent templates** for each image–trait pair. The total prompt count per evaluated model is therefore
$$
5{,}010 \times 100 \times 5 = 2{,}505{,}000
$$
(image, trait, template) combinations [2508.18989].

Bias is evaluated through the probability assigned to “Yes” for each prompt,
$$
P(\mathrm{Yes}\mid \mathrm{image},\mathrm{trait},\mathrm{template}),
$$
followed by between-group tests: Welch’s ANOVA for race, age, and skin tone, and Welch’s $t$-test for gender [2508.18989]. The paper introduces the **GRAS Bias Score**, defined as
$$
\mathrm{GRAS\ Bias\ Score}
=
\frac{1}{|A|\cdot |W|\cdot |T|}
\sum_{a\in A}\sum_{w\in W}\sum_{T=1}^{5}
\mathbb{I}(p_{a,w,T}>0.05)\times 100,
$$
so that the score is the percentage of attribute–trait–template cases in which no statistically significant between-group bias is detected. The scale is interpreted from **0** (maximal bias) to **100** (unbiased) [2508.18989].

The benchmarking results are severe. The abstract states that the least biased evaluated model attains a GRAS Bias Score of only **2 out of 100**, and the detailed table reports scores of **2.00** for *llava-1.5-7b-hf*, **1.75** for *paligemma2-3b-mix-224*, **1.00** for *Qwen2.5-VL-3B-Instruct*, **0.25** for *blip2-opt-2.7*, and **0.00** for *Phi-4-multimodal-instruct* [2508.18989]. The paper further reports that darker skin tones receive higher mean probabilities for more than 80% of negative traits, that lighter skin tones are more likely to be assigned positive traits, and that question formulation materially changes responses. A central methodological conclusion is that evaluating bias in VLMs with VQA requires considering multiple formulations of a question [2508.18989].

## 6. Disambiguation and cross-domain significance

A common source of confusion is that **GRAS** is not a single standardized research object. The same letter string, or a one-letter variant such as **GRASS**, identifies unrelated theories, conjectures, benchmarks, and models across mathematics, statistics, astrophysics, NLP, and fairness evaluation.

| Term | Referent | Representative paper |
|---|---|---|
| Gras conjecture | Relationship between unit-group indices and class groups in abelian extensions | [1102.0903], [1102.0617], [1502.01578] |
| GRASS | Graphical sure screening for Gaussian graphical models | [1407.7819] |
| GRAS | GRavitational Anti-Screening | [2602.09249] |
| Multi-GraS | Multiplex Graph Summarization for extractive summarization | [2108.12870] |
| GRAS | Gender, Race, Age, and Skin Tone benchmark for VLM bias | [2508.18989] |

The disciplinary separation is substantial. In number theory, the term is tied to unit groups, Stark units, Euler systems, class groups, and ray class groups [1102.0903]. In statistics, it refers to thresholding the sample covariance matrix to recover a conditional dependence graph with sure screening and false-positive-rate guarantees [1407.7819]. In astrophysics, it denotes a modified gravitational response that is mathematically equivalent to AQUAL in its standard form and modified to address cluster-scale discrepancies [2602.09249]. In machine learning, the same string appears in graph architectures for summarization and in a benchmark for demographic bias in VLMs [2108.12870]. This suggests that any technical use of “GRAS” should be read locally, with explicit attention to the surrounding field, notation, and cited literature.

Source: https://www.emergentmind.com/topics/gras