143174 research outputs found
Sort by
Meta-reasoning improves tool use in large language models
No full textExternal tools help large language models succeed at tasks where they would otherwise typically fail. In existing frameworks, choosing tools at test time relies on naive greedy decoding, regardless of whether the model has been fine-tuned on tool-annotated data or prompted with in-context examples. In contrast, we find that gathering and choosing among a suitable set of candidate tools has greater potential to lead to an optimal selection. We present Tool selECTion via meta-reasONing (TECTON), a two-phase system that first *reasons* over a task and outputs candidate tools using a custom fine-tuned language modelling head. Then, with the custom head disabled, it *meta-reasons* (i.e., it reasons over the previous reasoning process) to make a final choice. We show that TECTON results in substantial gains—both in-distribution and out-of-distribution—on a range of math reasoning datasets
Weight loss after obesity disrupts cognitive flexibility through reinforcement learning strategies
No full textObjective
Despite successful weight loss, many individuals with obesity regain weight, yet cognitive factors in the weight loss state remain unclear. Here, we tested whether obesity induces deficits in cognitive flexibility, a core component of reinforcement learning (RL), after body weight normalizes.
Methods
Male and female C57BL/6J mice were exposed to high-fat diet-induced obesity followed by weight loss. Weight loss and control mice were tested on a modified probabilistic reversal learning (PRL) task to assess cognitive flexibility and a progressive-ratio (PR) task to evaluate motivation. RL modeling was applied to dissociate latent decision-making parameters.
Results
Post-weight-loss mice exhibited persistent impairments in PRL efficiency. Males showed reduced late-phase reversal efficiency (p < 0.001), while females showed early-phase inefficiency but later recovery (p < 0.05). RL modeling revealed reduced learning rates in both sexes, indicating impaired value updating despite intact motivation, as PR performance did not differ between groups. Across tasks, food intake remained unchanged, suggesting reduced efficiency reflected cognitive inflexibility rather than diminished appetite.
Conclusions
Weight loss after obesity produced sex-specific RL deficits. These findings dissociated motivational drive from cognitive flexibility and highlighted maladaptive decision-making as a feature of the weight loss state. This demonstrates the need for targeted interventions addressing post-weight-loss cognitive barriers
The 3V score and joint associations of low ultra-processed food, biodiverse and plant-based diets on colorectal cancer risk: results from the European Prospective Investigation into Cancer and Nutrition (EPIC) Study
No full textBackground:
Diet may modify colorectal cancer risk. We investigated the associations of three dietary patterns, ultra-processed food (UPF) consumption, healthy plant-based food consumption, and food biodiversity, separately and combined into a “3V” score with risk of colorectal cancer.
Methods:
This study used data from the prospective European Prospective Investigation into Cancer and Nutrition (EPIC) cohort, which recruited participants between 1992, and 2000, from 23 centres in ten European countries. The 3V score was developed by standardising and summing the healthy plant diet index (hPDI) and dietary species richness per year (DSR) and subtracting UPF (Nova category 4) intake in % g/day. Associations with colorectal cancer risk were assessed among 450,111 middle-aged participants of the EPIC cohort using multivariable-adjusted Cox regression models. Independent associations of each 3V component were assessed using mutually adjusted models. Data-driven thresholds were applied to assess adherence to the 3V components, set at the minimum value of the fourth quintile for hPDI, DSR and low UPF.
Findings:
During mean (standard deviation (SD)) follow-up of 14.9 (4) years, absolute colorectal cancer rates were 8.59 and 10.37 cases/10,000 person-years for the highest and lowest quintiles of the 3V score, respectively. Inverse associations were found for colorectal (hazard ratio (HR) comparing highest versus lowest quintile: 0.84; 95% confidence interval (CI): 0.76–0.94), colon (HR: 0.82; 95% CI: 0.72–0.93), and distal colon cancer (HR: 0.81; 95% CI: 0.67–0.99), with significant linear trends observed across quintiles. UPF intake was positively associated with colon cancer risk (HR per 1 SD increment: 1.06; 95% CI: 1.02 –1.11) when mutually adjusted for the other 3V components. Adherence to low UPF, high hPDI, and high DSR was inversely associated with colorectal (HR: 0.73; 95% CI: 0.61–0.88), colon (HR: 0.72; 95% CI: 0.57–0.91), and rectal cancer (HR: 0.65; 95% CI: 0.46–0.91) compared to adhering to none.
Interpretation:
Adherence to the 3V diet is associated with lower risk of colorectal cancers.
Funding:
Cancer Research UK, World Cancer Research Fun
Numerical investigation of offshore foundation on liquefiable sands
No full textThis study investigates the seismic response of shallow foundations resting on liquefiable sand deposits, with the aim of providing insights into the expected behaviour of Wind Turbine Installation Vessels (WTIV) when subjected to earthquake loading. A detailed calibration strategy based on commonly available ground information is outlined for Nevada sand, with a detailed characterisation of the model performance being undertaken in terms of CSR, stiffness degradation, and damping ratio curves. Subsequent validation process is also provided by
simulating centrifuge experiments of footings resting on liquefiable deposits. Lastly, three-dimensional finite element analyses of a WTIV are performed employing the calibrated UBC3D-PLM parameters. The impact of soil liquefaction on the response of the WTIV is investigated, with particular emphasis given to the additional settlements caused by the seismic loading
Constructive proofs for some semilinear PDEs on H2(e|x|2/4, Rd)
No full textWe develop computer-assisted tools to study semilinear equations of the form
−Δu − x2· ∇u = f (x, u, ∇u), x ∈ Rd .
Such equations appear naturally in several contexts, and in particular when looking for
self-similar solutions of parabolic PDEs. We develop a general methodology, allowing
us not only to prove the existence of solutions, but also to describe them very precisely.
We introduce a spectral approach based on an eigenbasis of L := −Δ− x
2 ·∇ in spherical coordinates, together with a quadrature rule allowing to deal with nonlinearities,
in order to get accurate approximate solutions. We then use a Newton–Kantorovich
argument, in an appropriate weighted Sobolev space, to prove the existence of a
nearby exact solution. We apply our approach to nonlinear heat equations, to nonlinear
Schrödinger equations and to a generalised viscous Burgers equation, and obtain both
radial and non-radial self-similar profiles
Consistent and scalable monitoring of birds and habitats along a coffee production intensity gradient
No full textLand use change associated with agricultural intensification is a leading driver of biodiversity loss in the tropics. To evaluate the habitat–biodiversity relationship in production systems of tropical agricultural commodities, birds are commonly used as indicators. However, a consistent and reliable methodological approach for monitoring tropical avian communities and habitat quality in a way that is scalable is largely lacking. In this study, we examined whether the automated analysis of audio data collected by passive acoustic monitoring, together with the analysis of remote sensing data, can be used to efficiently monitor avian biodiversity along the gradient of habitat degradation associated with the intensification of coffee production. Coffee is an important crop produced in tropical forested regions, whose production is expanding and intensifying, and coffee production systems form a gradient of ecological complexity ranging from forest-like shaded polyculture to dense sun-exposed monoculture. We used LiDAR technology to survey the habitat, together with autonomous recording units and a vocalization classifier to assess bird community composition in a coffee landscape comprising a shade-grown coffee farm, a sun coffee farm and a forest remnant, located in southern Mexico. We found that LiDAR can capture relevant variation in vegetation across the habitat gradient in coffee systems, specifically matching the generally observed pattern that the intensification of coffee production is associated with a decrease in vegetation density and complexity. We also found that bioacoustics can capture known functional signatures of avian communities across this habitat degradation gradient. Thus, we show that these technologies can be used in a robust way to monitor how biodiversity responds to land use intensification in the tropics. A major advantage of this approach is that it has the potential to be deployed cost-effectively at large scales to help design and certify biodiversity-friendly productive landscapes
UPBEAT Trial: a randomised feasibility evaluation of a digital system for UPper limB rEhabilitation After sTroke
No full textA statistical method for evaluating vaccine-induced immune correlates of protection against infection and disease progression: application to the ChAdOx1-S nCoV-19 phase 3 trial
No full textBackground: Correlates of protection (CoPs), defined as immune markers statistically correlated with vaccine
efficacy (VE), can be used to accelerate vaccine development. Different components of the immune response may be important for protection against infection and against progression from asymptomatic infection to symptomatic or severe disease. However, CoPs are typically evaluated for these outcomes separately, which can lead to some CoPs not being identified. We propose a novel statistical framework for the integrated evaluation of CoPs for infections with multiple potential outcomes.
Methods: We developed a model of the natural history of an infection that can identify CoPs at each stage of
infection and disease progression and implemented this model in a Bayesian estimation framework. We validated
the model on simulated data then applied it to individual-level clinical and serum neutralising and binding
antibody data from COV002 (NCT04400838), a phase II/III trial of the ChAdOx1 nCoV-19 (AZD1222) vaccine.
We explored logistic and non-parametric (cubic spline) relationships between VE and the candidate CoPs.
Results: Both parametric and non-parametric forms of the model accurately estimated the relationships between the immune CoP and VE against infection (VEin) and against progression to symptoms given infection (VEpr) in 1000 simulated trial datasets. In the COV002 correlates subset (2227 participants, 5315 samples), SARS-CoV-2 spike-specific IgG was positively associated with both VEin and VEpr (average proportion of VE mediated by spike specific IgG, 27 % (95 % CI 2–88 %) for VEin and 41 % (95 % CI 0–96 %) for VEpr). Pseudoneutralisation antibody titres and receptor binding domain (RBD) specific serum IgG showed similar correlations.
Conclusion: Integrated analysis of multiple disease outcomes and candidate CoPs enables the identification of CoPs that operate at different stages of disease progression, which are missed when evaluating outcomes separately
The interpretation of triaxial test data and its effect on the numerical analysis of monopiles founded in sand
No full textThe PISA Joint Industry Project has demonstrated the need for advanced numerical analyses and the use of
sophisticated constitutive models in the assessment and design of monopile foundations for offshore wind turbine generators (WTGs). The predictions of any constitutive model depend on the parameters adopted to model the soil response, whilst the derivation of the model parameters depends on the interpretation of the laboratory and in situ tests carried out as part of the ground investigation for the offshore wind farm (OWF). This paper discusses the interpretation of triaxial test data on sands,
specifically their ultimate states, and the effect that this has on the calibration of two constitutive models and the predicted response of a monopile founded in sand deposits. The study employs two elasto-plastic constitutive models, i.e. a state parameter-dependent model (Taborda et al., 2018) and a strain-softening Mohr-Coulomb model (Potts et al., 1990) which have been previously calibrated against high-quality ground investigation information from an OWF (Grammatikopoulou et al., 2023). The paper examines key aspects of the monopile response, including load-displacement curves, soil reaction curves and structural forces, and highlights that different interpretations of the sands’ ultimate triaxial states can impact monopile design
