15064 research outputs found
Sort by
Explicitly unbiased large language models still form biased associations
Get PDFLarge language models (LLMs) can pass explicit social bias tests but still harbor implicit biases, similar to humans who endorse egalitarian beliefs yet exhibit subtle biases. Measuring such implicit biases can be a challenge: As LLMs become increasingly proprietary, it may not be possible to access their embeddings and apply existing bias measures; furthermore, implicit biases are primarily a concern if they affect the actual decisions that these systems make. We address both challenges by introducing two measures: LLM Word Association Test, a prompt-based method for revealing implicit bias; and LLM Relative Decision Test, a strategy to detect subtle discrimination in contextual decisions. Both measures are based on psychological research: LLM Word Association Test adapts the Implicit Association Test, widely used to study the automatic associations between concepts held in human minds; and LLM Relative Decision Test operationalizes psychological results indicating that relative evaluations between two candidates, not absolute evaluations assessing each independently, are more diagnostic of implicit biases. Using these measures, we found pervasive stereotype biases mirroring those in society in 8 value-aligned models across 4 social categories (race, gender, religion, health) in 21 stereotypes (such as race and criminality, race and weapons, gender and science, age and negativity). These prompt-based measures draw from psychology’s long history of research into measuring stereotypes based on purely observable behavior; they expose nuanced biases in proprietary value-aligned LLMs that appear unbiased according to standard benchmarks
Winter warming of McMurdo Dry Valleys soils
Get PDFContinuous permafrost is present across the McMurdo Dry Valleys of southern Victoria Land, Antarctica. While summer active-layer thaw is common in the low-elevation portions of the Dry Valleys, active layers have not significantly thickened over time. However, in some locations, coastal Antarctic permafrost has begun to warm. Here, based on soil and meteorological measurements from 1993 to 2023, we show that wintertime soil temperatures have increased across multiple sites in the Dry Valleys, at rates exceeding the pace of summer soil warming. Linear warming trends over time are significant (P < 0.05) at six of seven soil monitoring sites. Winter warming is strongly correlated with increased numbers of down-valley wind events (Foehn/katabatics), but it may also be driven by increased incident longwave radiation at some stations (although winter longwave increase is not significant over time). While down-valley wind events increase winter warming, when down-valley wind events are excluded from the record, winter soil warming remains persistent and significant, suggesting that Antarctic soils are experiencing less cold winters over time in response to regional warming. Together, these observations suggest that some Antarctic permafrost may be approaching a transition to discontinuous permafrost in some regions as winter freezing intensity is reduced over time
Multiscale simulation and machine learning facilitated design of two-dimensional nanomaterials-based tunnel field-effect transistors: A review
Get PDFTraditional transistors based on complementary metal–oxide–semiconductor and metal–oxide–semiconductor field-effect transistors are facing significant limitations as device scaling reaches the limits of Moore’s law. These limitations include increased leakage currents, pronounced short-channel effects, and quantum tunneling through the gate oxide, leading to higher power consumption and deviations from ideal behavior. Tunnel Field-Effect Transistors (TFETs) can overcome these challenges by utilizing the quantum tunneling of charge carriers to switch between on and off states and achieve a subthreshold swing below 60 mV/decade. This allows for lower power consumption, continued scaling, and improved performance in low-power applications. This review focuses on the design and operation of TFETs, emphasizing the optimization of device performance through material selection and advanced simulation techniques. The discussion will specifically address the use of two-dimensional materials in TFET design and explore simulation methods ranging from multi-scale approaches to machine learning-driven optimization
Reproducibility Report for ACM SIGMOD 2024 Paper: "Fast Maximal Quasi-clique Enumeration: A Pruning and Branching Co-Design Approach"
Get PDFThe core thesis of the paper has been successfully reproduced. The authors provided a set of scripts that can execute the experiments and generate the individual plots. The resulting figures were near identical to the submitted paper
Online Female Mutual Support Communities: Themes and Engagement Mechanism
No full textOnline mutual support communities provide vital spaces for individuals to seek guidance, share experiences, and engage in collective discourse. This study examines engagement dynamics within Douban’s \textit{Women in Academia} group, a digital support community for women in higher education. Using a large-scale dataset spanning over three years, this research employs Latent Dirichlet Allocation (LDA) to identify prevalent discussion topics and utilizes machine learning modeling to investigate factors influencing post engagement. The findings reveal that discussions primarily center around research advice, emotional expression, and feminist discourse, aligning with broader literature on online social support and gendered digital engagement. The results highlight that historical user activity, emotional intensity, and participatory behaviors strongly predict engagement levels. Additionally, interaction effects suggest that emotional expression plays a particularly significant role in feminist discussions. These insights contribute to the understanding of digital support mechanisms and provide practical recommendations for optimizing engagement strategies in online feminist and academic communities
Full genomes of all nine currently recognized lovebird species (genus <i>Agapornis</i>) sampled from wild populations
Get PDFAfrican lovebirds are popular pet parrots commonly hybridized by breeders leading to genetic admixture. Trade escapees and land use change have now led lovebirds to genetically admix in the wild. Sampling origin is therefore of utmost importance when deriving genetic data from lovebird species to reconstruct phylogeny or historical demographic events as the inclusion of taxa of hybrid origin is a source of spurious results. Here we present complete genomes of all nine currently recognized lovebird species. Each species is represented by an archival geo-referenced individual collected within their natural range
Asymptotic Notions of Computability
Get PDFIn computability theory, the classical definition of a Turing machine M solving a problem requires M to always halt and always give the right answer. If we relax the “always” to “almost always”, we give rise to asymptotic notions of computability, where now M halts and gives correct answers only asymptotically always. There are four asymptotic notions of computability: generic, coarse, dense, and effective dense computability. This dissertation studies these asymptotic notions of computability, specifically the degree structures arising from the associated reducibilities. Two of the main results concerns minimal pairs: pairs of sets which are both noncomputable, and which have no common computational power. We prove that there are only measure-0 many minimal pairs for the generic degrees (which is in contrast with the classic Turing degrees, for which there are measure-1 many minimal pairs) and construct a minimal pair for the coarse degrees. The third main result concerns attractive degrees. The formalization of the asymptotic notions of computability can be generalized to provide a notion of distance between Turing degrees, called the Hausdorff distance H. It turns out that for every set A, there are either measure-1 many sets which are at a distance 1 from A, or there are measure-1 many such sets which are at a distance 1/2 from A. The former are called dispersive, and the latter are called attractive. We provide a Kolmogorov-complexity flavored sufficient condition for a set to be attractive
Mechanistic Studies of HIV-1 Capsid Maturation and Nuclear Entry Using Multiscale Coarse-Grained Simulations
Get PDFDuring HIV-1 maturation, CA can self-assemble into a wide range of capsid morphologies made of ~175-250 hexamers and 12 pentamers. Most recently, the cellular polyanion inositol hexakisphosphate (IP6) has been demonstrated to facilitate conical capsid formation by coordinating a ring of arginine residues within the central cavity of capsid hexamers and pentamers. However, the precise kinetic interplay of events during IP6 and CA co-assembly is unclear. In the first project, we use Coarse-grained Molecular Dynamics (CGMD) simulations to elucidate the underlying molecular mechanism of capsid formation, including the crucial role played by IP6. We show that IP6, in relatively small quantities at first, promotes curvature generation by trapping pentameric defects in the growing lattice and shifts assembly behavior towards kinetically favored outcomes. Our analysis also suggests that IP6 can stabilize metastable capsid intermediates and can induce structural pleomorphism in mature capsids. Relatedly, A structural switch comprising the Thr-Val-Gly-Gly (TVGG) motif either assumes a disordered coil or a helix conformation to regulate hexamer or pentamer assembly, respectively. Both IP6 binding and TVGG coil-to-helix transition are essential for pentamer formation. However, the correlation between IP6 binding at the pore and mechanistic details of coil-to-helix transition in pentamer have not been elucidated. Using extensive all-atom molecular dynamics simulations and structural analysis, we demonstrate that IP6 binding at the pore triggers a network of interactions downstream. IP6 imparts structural order at the central ring, which results in multiple kinetically controlled events leading to the coil-to-helix conformational change of the TVGG motif. IP6 facilitates the helix-to-coil transition by allowing the formation of intermediate conformations. Our results identify the key kinetic role of IP6 in pentamer formation, which facilitates the capsid assembly. These results also may point to new druggable targets to prevent intact HIV-1 core formation. For example, small molecule Lenacapavir (LEN) has been proposed to disrupt capsid morphogenesis by occupying the FG-binding pocket located between neighboring CA subunits. As LEN and IP6 interact with overlapping structural elements, they can compete to influence the assembly pathway and outcomes. Using coarse-grained molecular simulations, we examined capsid assembly across varying IP6 and LEN conditions. Our results reveal a concentration-dependent shift in assembly outcomes: LEN accelerates hexamer assembly and reduces pentamer incorporation, leading to malformed, multilayered, or incomplete capsids. Simulations including the viral RNP further show that LEN-treated capsids frequently fail to encapsidate the RNA genome, indicating impaired maturation. Our calculations confirm that LEN impairs the formation of high-curvature regions necessary for closure, supporting a model of off-pathway assembly as a mechanism of viral inhibition. Building on these findings, we developed a bottom-up coarse-grained modeling framework to investigate the interaction of the HIV-1 capsid with host factors during nuclear entry. We constructed and simulated CG models of the capsid in complex with cyclophilin A, Nup358, and Nup153. Our results reveal that while moderate levels of CypA binding stabilize the capsid, excessive CypA coating induces asymmetric strain and structural collapse, particularly at the narrow tip. Simulations of Nup358 domains captured known structural features such as the S-shaped N-terminal solenoid and the oligomerization-driven filament formation via the OE domain. We further modeled Nup153-mediated interactions at the nuclear basket and observed multivalent FG-pocket binding and weak Nup153 self-association, suggesting a possible role in directional capsid translocation and mesh formation. These findings provide a unified view of HIV-1 capsid behavior across multiple stages of viral replication, from self-assembly to nuclear import. The integrative modeling approach presented here lays a foundation for future investigations into viral uncoating, host restriction and antiviral design targeting capsid-host interactions
Abolitionism and Ecosocial Work: Towards Equity, Liberation and Environmental Justice
Get PDFThe ecosocial work approach incorporates the environment into social justice frameworks. It calls for holistic practices that centre Indigenous and Global South voices, transdisciplinarity, anti-oppressive strategies in micro–macro practice and sustainability. In this article, we argue that the integration of abolitionist theory and practice within ecosocial work—specifically the reduction or elimination of social work from carceral systems that harm the environment and marginalised people—will better equip social workers to meet these goals. To this end, we critically examine three sites of ongoing abolitionist struggle—militaries, prisons and disinvested communities—to demonstrate reimagined strategies for an environmentally just future. Our analysis reveals areas where the ecosocial approach and abolitionism intersect, and how one might inform the other. We conclude by calling for future research to integrate abolitionist ideas into ecosocial work research, education and practice. In integrating environmental justice and abolition, ecosocial work will be better positioned to critique and resist its location in racialised capitalist systems that perpetuate economic, environmental, racial and social injustices
No anomalous supersaturation in ultracold cirrus laboratory experiments
Get PDFHigh-altitude cirrus clouds are climatically important: their formation freeze-dries air ascending to the stratosphere to its final value, and their radiative impact is disproportionately large. However, their formation and growth are not fully understood, and multiple in situ aircraft campaigns have observed frequent and persistent apparent water vapor supersaturations of 5 %–25 % in ultracold cirrus (T<205 K), even in the presence of ice particles. A variety of explanations for these observations have been put forth, including that ultracold cirrus are dominated by metastable ice whose vapor pressure exceeds that of hexagonal ice. The 2013 IsoCloud campaign at the Aerosol Interaction and Dynamics in the Atmosphere (AIDA) cloud and aerosol chamber allowed explicit testing of cirrus formation dynamics at these low temperatures. A series of 28 experiments allows robust estimation of the saturation vapor pressure over ice for temperatures between 189 and 235 K, with a variety of ice nucleating particles. Experiments are rapid enough (∼10 min) to allow detection of any metastable ice that may form, as the timescale for annealing to hexagonal ice is hours or longer over the whole experimental temperature range. We show that in all experiments, saturation vapor pressures are fully consistent with expected values for hexagonal ice and inconsistent with the highest values postulated for metastable ice, with no temperature-dependent deviations from expected saturation vapor pressure. If metastable ice forms in ultracold cirrus clouds, it appears to have a vapor pressure indistinguishable from that of hexagonal ice to within about 4.5 %