Birkbeck Institutional Research Online

Birkbeck, University of London

Birkbeck Institutional Research Online
Not a member yet
    30472 research outputs found

    A Seat at the Table: A Place for Disabled Entrepreneurs and Innovators

    No full text
    This edited book addresses the question of how the innovation and entrepreneurship created by disabled people is supported in different national contexts. It explores factors which influence the availability of resources in support of entrepreneurship and innovation by disabled people. Country case studies in this volume are India, Nigeria, Hungary, the USA, and Türkiye. The book also draws on previous research in the UK

    Anchor attention, small cache: code generation with Large Language Models

    Get PDF
    The development of large language models (LLMs) has revolutionized automated code generation. However, their high demand of computation resources has hindered a broader deployment and raised environmental concerns. A common strategy for diminishing computational demands is to cache Key-Value (KV) states from the attention mechanism which is adopted predominately by mainstream LLMs. It can mitigate the need of repeated attention computations, but brings significant memory overhead. Current practices in NLP often use sparse attention which may, unfortunately, lead to substantial inaccuracies, or hallucinations, in code generation tasks. In this paper, we analyze the attention weights distribution within code generation models via an empirical study, uncovering a sparsity pattern, i.e., the aggregation of information at specific anchor points. Based on this observation, we propose a novel approach, AnchorCoder, which features token-wise anchor attention designed to extract and compress the contextual information, and layer-wise anchor attention enabling cross-layer communication to mitigate the issue of excessive superposition caused by the compression. The extensive experiments across multiple benchmark datasets confirm the effectiveness of AnchorCoder, which can consistently achieve a significant (at least 70%) reduction in KV cache requirements, while preserving the majority of model’s performance

    Interstellar migration, the population ‘Problem’ and the climate emergency

    Get PDF
    A decade after its initial release, this article explores the significance of Christopher Nolan’s epic film, Interstellar (2014), and suggests that it provides the opportunity of reflecting on some of the key questions of our age: is techno-optimism and its ultimate expression - invocations of interstellar migration - a byword for climate denial, an anti-ecological refusal of natural limits? What happens to environmentalist thinking when it absorbs human population into its understanding of what those ‘natural’ limits are? And what are our responsibilities to future generations in the context of the climate emergency? With space travel back on the international agenda and a new Trump administration preaching the need for Americans to procreate, this piece deploys Interstellar as a way of exploring these intertwined issues of population, climate and technology

    Unbiased and error-detecting combinatorial pooling experiments with balanced constant-weight gray codes for consecutive positives detection

    Get PDF
    Motivation: Combinatorial pooling schemes have enabled the measurement of thousands of experiments in a small number of reactions. This efficiency is achieved by distributing the items to be measured across multiple reaction units called pools. However, current methods for the design of pooling schemes do not adequately address the need for balanced item distribution across pools, a property particularly important for biological applications. Results: Here, we introduce balanced constant-weight Gray codes for detecting consecutive positives (DCP-CWGCs) for the efficient construction of combinatorial pooling schemes. Balanced DCP-CWGCs ensure uniform item distribution across pools, allow for the identification of consecutive positive items such as overlapping biological sequences, and enable error detection by ensuring a constant number of tests on each item and pair of consecutive items. For the efficient construction of balanced DCP-CWGCs, we have released an open-source python package codePUB, with implementations of the two core algorithms: a branch-and-bound algorithm (BBA) and a recursive combination with BBA (rcBBA). Simulations using codePUB show that our algorithms can construct long, balanced DCP-CWGCs that allow for error detection in tractable runtime. Availability and implementation: The source code of codePUB is available at https://github.com/meyer-lab-cshl/ codepub, with detailed documentation at https://codepub.readthedocs.io/

    SEE+ computerized classroom-based training enhances 7- to 10-year-olds' socio-emotional cognition through observation and inference

    Get PDF
    Smooth social interactions rely on children’s abilities to decode others’ social signals, which includes what an individual may say or do, and their facial emotional expressions. Failure can lead to exclusion from social groups. Consequently, a number of social and emotional learning (SEL) training programmes have been developed, with some evidence of positive impacts on those skills themselves and on academic achievement. Here, we present and evaluate the efficacy of a novel classroom-based computerized learning activity, called SEE+ (Socio-emotional engagement through observation), in supporting and enhancing 7- to 10-year-olds’ emotion recognition and theory of mind judgements through observation and inference. SEE+ involved observing four virtual characters interacting within social scenarios and inferring their mental states. Participants were recruited from across diverse school settings (rural, urban) in England as part of a large-scale randomised controlled trial (n = 5585; 7.3–11.0 years old, M age = 9.1, SD = 1.0) resulting in a mix of socio-economic and ethnic backgrounds. Mixed model ANOVAs were used to compare performance on socio-emotional tasks between the SEE+ group, an active control group and a teaching-as-usual control group. SEE + was associated with the equivalent of 4–6 months improvement in performance on the socio-emotional assessment tasks. Children showed near transfer effects of the intervention (i.e., when characters present in the computerized learning activity were used in the assessment), while no strong evidence of far transfer effects (i.e., when photographs of unknown children were used in the assessment) was found. Limitations were the use of pen-and-paper assessments with a reduced number of trials, and a possible ceiling effect in the older children in the photograph condition. Our findings point to the relative plasticity of younger children’s socio-emotional cognition and underscore SEE+ as an easy-to-use cost-effective socio-emotional resource for teachers to embed SEL in the school curriculum

    No room at the inn? The case for dedicated replication journals

    No full text
    Replication is widely recognized as essential for scientific self-correction, yet published replications remain exceedingly rare in both economics and psychology. At the same time, large-scale collaborative replication efforts and growing attention to research transparency have revealed widespread reproducibility failures across empirical research. This article reviews evidence on the scarcity of published replications, examines why traditional journals continue to resist publishing them, and evaluates the “first-best” proposal that journals should publish replications of their own articles. It then surveys alternative models that journals have adopted, such as dedicated replication sections and special replication issues, and considers their limitations in overcoming structural barriers to replication. Because these approaches have not meaningfully expanded the publication of replications, the article argues that dedicated replication journals offer an essential and complementary solution. By providing a stable, credible, and visible home for replication work, these journals supply critical infrastructure for the self-correcting function of science

    Revisiting the “Surveillant Assemblage”

    Get PDF
    The “surveillant assemblage” and its attendant notion of the “data double” have held a central place within surveillance theory for nearly a quarter of a century. Haggerty and Ericson proposed the concept as a way to push back against prevailing, and limited, surveillance theory rooted in Foucauldian panopticism and Orwellian notions of “big brother.” Yet, in an informational and technological landscape that has changed significantly since 2000, and continues to change rapidly, it seems vital to ask whether a concept like the surveillant assemblage continues to have meaningful application to surveillance today. In this Dialogue, we bring together surveillance scholars from across the humanities and social sciences to look at how the “surveillant assemblage” has contributed to surveillance scholarship and informed our thinking about surveillance in society. The Dialogue includes six interventions from scholars who apply, extend, refashion, and critique the surveillant assemblage, as well as a closing response from Kevin Haggerty—the first time he has chosen to directly revisit the idea of the surveillant assemblage. In his response, Haggerty outlines a proposal for shifting our attention to understanding surveillance as a process of “surveillantization.

    Handling 403 errors in BuddyPress's new_activity_comment

    No full text
    A technical post on BuddyPres

    On academic freedom, provisional whiteness and Antisemitism

    No full text
    In the furore over Derek Hook’s article White Anxiety and Donald Moss’ On Having Whiteness, an under-reported but important issue is how easily and strongly antisemitism surfaced. Derek Hook is not Jewish but was assumed to be so by some of his abusive critics on the grounds probably that no white person would make the kinds of statements he was alleged to make (themselves of course a distortion of his actual article). Moss is Jewish, which makes it easier for his critics to see him as an adversary; not so much a ‘race traitor’ as a Jew who has relinquished his provisional whiteness. In both cases, one issue amongst the many to do with ‘academic freedom’ is how and why the racialisation of hatred can occur so strongly, often as antisemitism, in the context of psychoanalytic explorations of racism and especially of whiteness

    10,887

    full texts

    30,472

    metadata records
    Updated in last 30 days.
    Birkbeck Institutional Research Online is based in United Kingdom
    Access Repository Dashboard
    Do you manage Birkbeck Institutional Research Online? Access insider analytics, issue reports and manage access to outputs from your repository in the CORE Repository Dashboard!