搜索结果：associations

Human word associations are a well-known method of gaining insight into the internal mental lexicon, but the responses spontaneously offered by human participants to word cues are not always predictable as they may be influenced by personal experience, emotions or individual cognitive styles. The ability to form associative links between seemingly unrelated concepts can be the driving mechanisms of creativity. We perform a comparison of the associative behaviour of humans compared to large language models. More specifically, we explore associations to emotionally loaded words and try to determine whether large language models generate associations in a similar way to humans. We find that the overlap between humans and LLMs is moderate, but also that the associations of LLMs tend to amplify the underlying emotional load of the stimulus, and that they tend to be more predictable and less creative than human ones.

OB associations are important probes of recent star formation and Galactic structure. In this study, we focus on the Auriga constellation, an important region of star formation due to its numerous young stars, star-forming regions and open clusters. We show using \textit{Gaia} data that its two previously documented OB associations, Aur OB1 and OB2, are too extended in proper motion and distance to be genuine associations, encouraging us to revisit the census of OB associations in Auriga with modern techniques. We identify 5617 candidate OB stars across the region using photometry, astrometry and our SED fitting code, grouping these into 5 high-confidence OB associations using HDBSCAN. Three of these are replacements to the historical pair of associations - Aur OB2 is divided between a foreground and a background association - while the other two associations are completely new. We connect these OB associations to the surrounding open clusters and star-forming regions, analyse them physically and kinematically, constraining their ages through a combination of 3D kinematic traceback, the position of their members in the HR diagram and their connection to clusters of known age. Four

Existing works examining Vision-Language Models (VLMs) for social biases predominantly focus on a limited set of documented bias associations, such as gender:profession or race:crime. This narrow scope often overlooks a vast range of unexamined implicit associations, restricting the identification and, hence, mitigation of such biases. We address this gap by probing VLMs to (1) uncover hidden, implicit associations across 9 bias dimensions. We systematically explore diverse input and output modalities and (2) demonstrate how biased associations vary in their negativity, toxicity, and extremity. Our work (3) identifies subtle and extreme biases that are typically not recognized by existing methodologies. We make the Dataset of retrieved associations, (Dora), publicly available here https://github.com/chahatraj/BiasDora.

Despite the remarkable performance of foundation vision-language models, the shared representation space for text and vision can also encode harmful label associations detrimental to fairness. While prior work has uncovered bias in vision-language models' (VLMs) classification performance across geography, work has been limited along the important axis of harmful label associations due to a lack of rich, labeled data. In this work, we investigate harmful label associations in the recently released Casual Conversations datasets containing more than 70,000 videos. We study bias in the frequency of harmful label associations across self-provided labels for age, gender, apparent skin tone, and physical adornments across several leading VLMs. We find that VLMs are $4-7$x more likely to harmfully classify individuals with darker skin tones. We also find scaling transformer encoder model size leads to higher confidence in harmful predictions. Finally, we find improvements on standard vision tasks across VLMs does not address disparities in harmful label associations.

Dense retrieval systems rank passages by embedding similarity to a query, but multi-hop questions require passages that are associatively related through shared reasoning chains. We introduce Association-Augmented Retrieval (AAR), a lightweight transductive reranking method that trains a small MLP (4.2M parameters) to learn associative relationships between passages in embedding space using contrastive learning on co-occurrence annotations. At inference time, AAR reranks an initial dense retrieval candidate set using bi-directional association scoring. On HotpotQA, AAR improves passage Recall@5 from 0.831 to 0.916 (+8.6 points) without evaluation-set tuning, with gains concentrated on hard questions where the dense baseline fails (+28.5 points). On MuSiQue, AAR achieves +10.1 points in the transductive setting. An inductive model trained on training-split associations and evaluated on unseen validation associations shows no significant improvement, suggesting that the method captures corpus-specific co-occurrences rather than transferable patterns. Ablation studies support this interpretation: training on semantically similar but non-associated passage pairs degrades retrieval belo

In recent decades, traditional drug research and development have been facing challenges such as high cost, long timelines, and high risks. To address these issues, many computational approaches have been suggested for predicting the relationship between drugs and diseases through drug repositioning, aiming to reduce the cost, development cycle, and risks associated with developing new drugs. Researchers have explored different computational methods to predict drug-disease associations, including drug side effects-disease associations, drug-target associations, and miRNAdisease associations. In this comprehensive review, we focus on recent advances in predicting drug-disease association methods for drug repositioning. We first categorize these methods into several groups, including neural network-based algorithms, matrixbased algorithms, recommendation algorithms, link-based reasoning algorithms, and text mining and semantic reasoning. Then, we compare the prediction performance of existing drug-disease association prediction algorithms. Lastly, we delve into the present challenges and future prospects concerning drug-disease associations.

OB associations play an important role in Galactic evolution, though their origins and dynamics remain poorly studied, with only a small number of systems analysed in detail. In this paper we revisit the existence and membership of the Cygnus OB associations. We find that of the historical OB associations only Cyg OB2 and OB3 stand out as real groups. We search for new OB stars using a combination of photometry, astrometry, evolutionary models and an SED fitting process, identifying 4680 probable OB stars with a reliability of $>$90\%. From this sample we search for OB associations using a new and flexible clustering technique, identifying 6 new OB associations. Two of these are similar to the associations Cyg OB2 and OB3, though the others bear no relationship to any existing systems. We characterize the properties of the new associations, including their velocity dispersions and total stellar masses, all of which are consistent with typical values for OB associations. We search for evidence of expansion and find that all are expanding, albeit anistropically, with stronger and more significant expansion in the direction of Galactic longitude. We also identify two large-scale (1

We study the properties of associations of dwarf galaxies and their dependence on the environment. Associations of dwarf galaxies are extended systems composed exclusively of dwarf galaxies, considering as dwarf galaxies those galaxies less massive than $M_{\star, \rm max} = 10^{9.0}$ ${\rm M}_{\odot}\,h^{-1}$. We identify these particular systems using a semi-analytical model of galaxy formation coupled to a dark matter only simulation in the $Λ$ Cold Dark Matter cosmological model. To classify the environment, we estimate eigenvalues from the tidal field of the dark matter particle distribution of the simulation. We find that the majority, two thirds, of associations are located in filaments ($ \sim 67$ per cent), followed by walls ($ \sim 26 $ per cent), while only a small fraction of them are in knots ($ \sim 6 $ per cent) and voids ($ \sim 1 $ per cent). Associations located in more dense environments present significantly higher velocity dispersion than those located in less dense environments, evidencing that the environment plays a fundamental role in their dynamical properties. However, this connection between velocity dispersion and the environment depends exclusively on

We develop a method to identify and determine the physical properties of stellar associations using Hubble Space Telescope (HST) NUV-U-B-V-I imaging of nearby galaxies from the PHANGS-HST survey. We apply a watershed algorithm to density maps constructed from point source catalogues Gaussian smoothed to multiple physical scales from 8 to 64 pc. We develop our method on two galaxies that span the distance range in the PHANGS-HST sample: NGC 3351 (10 Mpc), NGC 1566 (18 Mpc). We test our algorithm with different parameters such as the choice of detection band for the point source catalogue (NUV or V), source density image filtering methods, and absolute magnitude limits. We characterise the properties of the resulting multi-scale associations, including sizes, number of tracer stars, number of associations, photometry, as well as ages, masses, and reddening from Spectral Energy Distribution fitting. Our method successfully identifies structures that occupy loci in the UBVI colour-colour diagram consistent with previously published catalogues of clusters and associations. The median ages of the associations increases from log(age/yr) = 6.6 to log(age/yr) = 6.9 as the spatial scale incr

Transcriptome-wide association studies (TWAS) are powerful tools for identifying gene-level associations by integrating genome-wide association studies and gene expression data. However, most TWAS methods focus on linear associations between genes and traits, ignoring the complex nonlinear relationships that may be present in biological systems. To address this limitation, we propose a novel framework, QTWAS, which integrates a quantile-based gene expression model into the TWAS model, allowing for the discovery of nonlinear and heterogeneous gene-trait associations. Via comprehensive simulations and applications to both continuous and binary traits, we demonstrate that the proposed model is more powerful than conventional TWAS in identifying gene-trait associations.

The first gravitational wave (GW) - gamma-ray burst (GRB) association, GW170817/GRB 170817A, had an offset in time, with the GRB trigger time delayed by $\sim$1.7 s with respect to the merger time of the GW signal. We generally discuss the astrophysical origin of the delay time, $Δt$, of GW-GRB associations within the context of compact binary coalescence (CBC) -- short GRB (sGRB) associations and GW burst -- long GRB (lGRB) associations. In general, the delay time should include three terms, the time to launch a clean (relativistic) jet, $Δt_{\rm jet}$; the time for the jet to break out from the surrounding medium, $Δt_{\rm bo}$; and the time for the jet to reach the energy dissipation and GRB emission site, $Δt_{\rm GRB}$. For CBC-sGRB associations, $Δt_{\rm jet}$ and $Δt_{\rm bo}$ are correlated, and the final delay can be from 10 ms to a few seconds. For GWB-lGRB associations, $Δt_{\rm jet}$ and $Δt_{\rm bo}$ are independent. The latter is at least $\sim$10 s, so that $Δt$ of these associations is at least this long. For certain jet launching mechanisms of lGRBs, $Δt$ can be minutes or even hours long due to the extended engine waiting time $Δt_{\rm wait}$ to launch a jet. We d

We analyze the storage and recall of factual associations in autoregressive transformer language models, finding evidence that these associations correspond to localized, directly-editable computations. We first develop a causal intervention for identifying neuron activations that are decisive in a model's factual predictions. This reveals a distinct set of steps in middle-layer feed-forward modules that mediate factual predictions while processing subject tokens. To test our hypothesis that these computations correspond to factual association recall, we modify feed-forward weights to update specific factual associations using Rank-One Model Editing (ROME). We find that ROME is effective on a standard zero-shot relation extraction (zsRE) model-editing task, comparable to existing methods. To perform a more sensitive evaluation, we also evaluate ROME on a new dataset of counterfactual assertions, on which it simultaneously maintains both specificity and generalization, whereas other methods sacrifice one or another. Our results confirm an important role for mid-layer feed-forward modules in storing factual associations and suggest that direct manipulation of computational mechanisms

Quasar-galaxy associations, if they result from the effect of gravitational lensing by foreground galaxies, depend sensitively on the shape of the quasar number counts. Two kinds of quasar number-magnitude relations are predicted to produce quite different properties in quasar-galaxy associations: the counts of Boyle, Shanks and Peterson (1988; BSP) provide both positive and ``negative" associations between distant quasars and foreground galaxies, relating closely with the knee ($B\approx19.15$) in these counts. However, Hawkins and Véron (1993; HV) quasar data lead to only a positive magnitude-independent quasar-galaxy association. The current observational evidence on quasar-galaxy associations, either positive or null, is shown to be the natural result of gravitational lensing if quasars follow the BSP number-magnitude relation. On the other hand, the HV counts are unable to produce the reported associations by the mechanism of gravitational lensing. It is emphasized that special attention should be paid to the limiting magnitudes in the selected quasar samples when one works on quasar-galaxy associations.

The Milky Way's most luminous, young and massive (M > 10^4 Msun) star clusters and OB associations have largely evaded detection despite knowledge of their surrounding H II regions. We search for these clusters and associations within the 40 star forming complexes from Rahman & Murray in the 13 most luminous WMAP free-free emission sources of the Galaxy. Selecting for objects with the dust-reddened colors of OB stars, we identify new candidate associations using the 2MASS point source catalog. In 40 star forming complexes searched, 22 contain cluster/association candidates with sizes and masses in the range of 3' - 26' and 10^{2.3} - 10^{5} Msun. Of the 22 candidates, at least 7 have estimated masses > 10^4 Msun, doubling the number of such massive clusters known in the Galaxy. Applying our method to a statistically similar set of test locations, we estimate that 3.0 +/- 0.6 of our 22 candidate associations are unrelated to the star forming complexes. In addition, the apparent extinctions of our candidate associations correlate well with the predictions from a Galactic model. These facts, along with the clear detection of a known OB association and the previous spectral v

We investigate the associations between background galaxies and foreground clusters of galaxies due to the effect of gravitational lensing by clusters of galaxies. Similar to the well-known quasar-galaxy ones, these associations depend sensitively on the shape of galaxy number-magnitude or number-flux relation, and both positive and ``negative" associations are found to be possible, depending on the limiting magnitude and/or the flux threshold in the surveys. We calculate the enhancement factors assuming a singular isothermal sphere model for clusters of galaxies and a pointlike model for background sources selected in three different wavelengths, $B$, $K$ and radio. Our results show that $K-$ selected galaxies might constitute the best sample to test the ``negative" associations while it is unlikely that one can actually observe any association for blue galaxies. We also point out that bright radio sources ($S>1$ Jy) can provide strong positive associations, which may have been already detected in 3CR sample.

We have carried out a study of the early type stars in nearby OB associations spanning an age range of $\sim$ 3 to 16 Myr, with the aim of determining the fraction of stars which belong to the Herbig Ae/Be class. We studied the B, A, and F stars in the nearby ($\le 500$ pc) OB associations Upper Scorpius, Perseus OB2, Lacerta OB1, and Orion OB1, with membership determined from Hipparcos data. We obtained spectra for 440 Hipparcos stars in these associations, from which we determined accurate spectral types, visual extinctions, effective temperatures, luminosities and masses, using Hipparcos photometry. Using colors corrected for reddening, we find that the Herbig Ae/Be stars and the Classical Be stars (CBe) occupy clearly different regions in the JHK diagram. Thus, we use the location on the JHK diagram, as well as the presence of emission lines and of strong 12 microns flux relative to the visual to identify the Herbig Ae/Be stars in the associations. We find that the Herbig Ae/Be stars constitute a small fraction of the early type stellar population even in the younger associations. Comparing the data from associations with different ages and assuming that the near-infrared exces

A methodology is proposed to automatically detect significant symbol associations in genomic databases. A new statistical test is proposed to assess the significance of a group of symbols when found in several genesets of a given database. Applied to symbol pairs, the thresholded p-values of the test define a graph structure on the set of symbols. The cliques of that graph are significant symbol associations, linked to a set of genesets where they can be found. The method can be applied to any database, and is illustrated MSigDB C2 database. Many of the symbol associations detected in C2 or in non-specific selections did correspond to already known interactions. On more specific selections of C2, many previously unkown symbol associations have been detected. These associations unveal new candidates for gene or protein interactions, needing further investigation for biological evidence.

Low-mass stars 0.1 ~< M ~< 1 Msun) in OB associations are key to addressing some of the most fundamental problems in star formation. The low-mass stellar populations of OB associations provide a snapshot of the fossil star-formation record of giant molecular cloud complexes. Large scale surveys have identified hundreds of members of nearby OB associations, and revealed that low-mass stars exist wherever high-mass stars have recently formed. The spatial distribution of low-mass members of OB associations demonstrate the existence of significant substructure ("subgroups"). This "discretized" sequence of stellar groups is consistent with an origin in short-lived parent molecular clouds within a Giant Molecular Cloud Complex. The low-mass population in each subgroup within an OB association exhibits little evidence for significant age spreads on time scales of ~10 Myr or greater, in agreement with a scenario of rapid star formation and cloud dissipation. The Initial Mass Function (IMF) of the stellar populations in OB associations in the mass range 0.1 ~< M ~< 1 Msun is largely consistent with the field IMF, and most low-mass pre-main sequence stars in the solar vicinity ar

Gene-based testing is a commonly employed strategy in many genetic association studies. Gene-trait associations can be complex due to underlying population heterogeneity, gene-environment interactions, and various other reasons. Existing gene-based tests, such as Burden and Sequence Kernel Association Tests (SKAT), are based on detecting differences in a single summary statistic, such as the mean or the variance, and may miss or underestimate higher-order associations that could be scientifically interesting. In this paper, we propose a new family of gene-level association tests which integrate quantile rank score processes to better accommodate complex associations. The resulting test statistics have multiple advantages: (1) they are almost as efficient as the best existing tests when the associations are homogeneous across quantile levels, and have improved efficiency for complex and heterogeneous associations, (2) they provide useful insights on risk stratification, (3) the test statistics are distribution-free, and could hence accommodate a wide range of underlying distributions, and (4) they are computationally efficient. We established the asymptotic properties of the propose