Time-Aware Language Models as Temporal Knowledge Bases Dhingra, Bhuwan; Cole, Jeremy R.; Eisenschlos, Julian Martin ...
Transactions of the Association for Computational Linguistics,
03/2022, Letnik:
10
Journal Article
Recenzirano
Odprti dostop
Many facts come with an expiration date, from the name of the President to the basketball team Lebron James plays for. However, most language models (LMs) are trained on snapshots of data collected ...at a specific moment in time. This can limit their utility, especially in the closed-book setting where the pretraining corpus must contain the facts the model should memorize. We introduce a diagnostic dataset aimed at probing LMs for factual knowledge that changes over time and highlight problems with LMs at either end of the spectrum—those trained on specific slices of temporal data, as well as those trained on a wide range of temporal data. To mitigate these problems, we propose a simple technique for jointly modeling text with its timestamp. This improves memorization of seen facts from the training time period, as well as calibration on predictions about unseen facts from future time periods. We also show that models trained with temporal context can be efficiently “refreshed” as new data arrives, without the need for retraining from scratch.
Disaster robotics is a growing field that is concerned with the design and development of robots for disaster response and disaster recovery. These robots assist first responders by performing tasks ...that are impractical or impossible for humans. Unfortunately, current disaster robots usually lack the maneuverability to efficiently traverse these areas, which often necessitate extreme navigational capabilities, such as centimeter-scale clearance. Recent work has shown that it is possible to control the locomotion of insects such as the Madagascar hissing cockroach (
) through bioelectrical stimulation of their neuro-mechanical system. This provides access to a novel agent that can traverse areas that are inaccessible to traditional robots. In this paper, we present a data-driven inertial navigation system that is capable of localizing cockroaches in areas where GPS is not available. We pose the navigation problem as a two-point boundary-value problem where the goal is to reconstruct a cockroach's trajectory between the starting and ending states, which are assumed to be known. We validated our technique using nine trials that were conducted in a circular arena using a biobotic agent equipped with a thorax-mounted, low-cost inertial measurement unit. Results show that we can achieve centimeter-level accuracy. This is accomplished by estimating the cockroach's velocity-using regression models that have been trained to estimate the speed and heading from the inertial signals themselves-and solving an optimization problem so that the boundary-value constraints are satisfied.
Physical inspection and sorting of foraminifera is a necessity in many research labs, as foraminifera serve as paleoenvironmental and chronostratigraphic indicators. In order to gain counts of ...species from samples, analyze chemical compositions, or extract morphological properties of foraminifera, research labs require human time and effort handling and sorting these microscopic fossils. The presented work describes Forabot, an open‐source system which can physically manipulate individual foraminifera for imaging and isolation with minimal human interaction. The major components to build a Forabot are outlined in this work, with supplementary information available which allows for other researchers to build a Forabot with low‐cost, off‐the‐shelf components. From a washed and sieved sample of hundreds of foraminifera, the Forabot is shown to be capable of isolating and imaging individual forams. The timing of the Forabot's current pipeline allows for the processing of up to 27 foram specimens per hour, a rate that can be improved for future classification purposes by reducing image quality and/or quantity. Along with the physical descriptions, the image processing and classification pipelines are also reviewed. A proof‐of‐concept classifier utilizes a finetuned VGG‐16 network to achieve a classification accuracy of 79% on a validation set of foraminifera images collected with Forabot. In conclusion, the system is able to be built by researchers for a low cost, effectively manipulate foraminifera with few mistakes, provide quality images for future research, and classify the species of imaged forams.
Plain Language Summary
Foraminifera or “forams” are abundant microscopic organisms found in the ocean, and their shells are a common component of seafloor mud. Mud cores can be used to understand ancient ocean conditions, and the types and chemistry of forams in a sample are useful environmental indicators. However, separating different types of forams is slow and tedious work which requires considerable expertise. We have designed and built a robot called Forabot, which picks up individual shells, takes high‐quality photographs of them, and moves them to a bin for sorting. We describe the system so that other researchers can build their own Forabot at low cost. The current version of Forabot is optimized for high‐quality imaging and is therefore relatively slow, but if it is instead used for classifying and sorting shells into different types, it can be optimized for speed. We discuss the preliminary performance of a classifier based on artificial intelligence, with overall accuracy of 79%. In conclusion, our robot can be built by researchers for a low cost, effectively manipulate forams with few mistakes, provide quality images for future research, and accurately classify the type of foram.
Key Points
The Forabot is a low‐cost and open‐source system for automated isolation and imaging of foraminifera
We explore Forabot performance as a physical manipulation tool
Forabot images are classified using deep learning with promising results for future integration
Introduction
Annual influenza vaccinations are recommended for adolescents and adults with moderate to severe asthma. This study investigated the effect of tezepelumab, a human monoclonal antibody ...that blocks the activity of thymic stromal lymphopoietin, on the humoral immune response to the quadrivalent seasonal influenza vaccine in patients with moderate to severe asthma.
Methods
VECTOR was a phase 3b, randomized, multicenter, double-blind, parallel-group, placebo-controlled study. Adolescents (aged 12–17 years) and young adults (aged 18–21 years) with moderate to severe asthma were enrolled across 15 centers in the USA. Patients received tezepelumab 210 mg or placebo subcutaneously at weeks 0, 4, 8, and 12, and a single dose of inactivated quadrivalent seasonal influenza vaccine at week 12 before receiving study treatment. Immediately before vaccination and at 4 weeks postvaccination (week 16), strain-specific antibody responses were assessed for four influenza antigens by hemagglutination inhibition (HAI) and microneutralization (MN) assays. Safety was assessed.
Results
Seventy patients were randomized to tezepelumab (
n
= 35) or placebo (
n
= 35). There were no meaningful differences in HAI or MN antibody responses between treatment groups at week 16. HAI assay geometric mean fold rises (GMFRs) for influenza strains were 1.76–7.34 for tezepelumab and 1.46–4.75 for placebo. MN assay GMFRs were 4.00–14.56 for tezepelumab and 3.56–10.62 for placebo. In the HAI assay, a fourfold or larger rise in antibody titer from weeks 12 to 16 occurred in 15.2–78.8% and 15.2–51.5% of tezepelumab and placebo recipients, respectively, and 97.0–100% of patients in both treatment groups achieved an antibody titer of at least 40 at week 16. No unexpected safety findings occurred.
Conclusion
There was no observed suppression of the humoral immune response after influenza vaccination in adolescents and young adults with moderate to severe asthma treated with tezepelumab. Therefore, the influenza vaccine can be administered to this patient population during tezepelumab treatment.
ClinicalTrials.gov identifier
NCT05062759
Heat and mass balance calculations are important for monitoring volcanoes with heated crater lakes, but for these lakes the critical process of evaporation can be substantially affected by the lake's ...influence on the air mass above it. Measurements in 2010 using a weather station on a buoy floating in Ruapehu Crater Lake enabled us to derive a relation between wind velocity above the lake and that measured at nearby weather stations, as well as providing direct evidence of the effect of the warm lake on the air above it. This supported the use of evaporation equations that allowed for the changing air conditions as incoming air became warmer and wetter from interaction with the lake, so decreasing the overall evaporation rate.
Heat and mass balance calculations using these parameters and equations during the period 2003 to early 2007, as Crater Lake filled before it overflowed, confirm the previously observed high ratio of total heat flow to steam volume.
► We recorded the atmospheric conditions above the warm Ruapehu Crater Lake. ► The air above the lake was hotter and wetter than ambient. ► Evaporation from warm lakes is affected by the changing air properties.
In this randomized trial, dupilumab resulted in a lower annualized rate of exacerbations than placebo among patients with COPD and an elevated blood eosinophil count.
This paper examines the effects of working memory size in incremental grammatical encoding during language production. Our experiment tests different variants of a computational-cognitive model that ...combines an empirically validated framework of general cognition, ACT-R, with a linguistic theory, Combinatory Categorial Grammar. The model is induced from a corpus of spoken dialogue. This methodology facilitates comparison of different strategies and working memory capacities according to the similarity of the model’s produced sentences to the corpus sentences. The experiment presented shows that while having more working memory available improves performance, using less working memory during realization does as well, even after controlling sentence length. Sentences realized with a more incremental strategy also appear to more closely track the naturalistic data. As high incrementality is correlated with low working memory usage, this study offers a possible mechanism by which syntactic incrementality can be explained. Finally, this paper proposes a multi-disciplinary modeling and simulation-based approach to empirical psycholinguistic inquiry.
Weight reduction is essential for improving health outcomes in people with obesity and type 2 diabetes. We assessed the efficacy and safety of tirzepatide, a glucose-dependent insulinotropic ...polypeptide and glucagon-like peptide-1 receptor agonist, versus placebo, for weight management in people living with obesity and type 2 diabetes.
This phase 3, double-blind, randomised, placebo-controlled trial was conducted in seven countries. Adults (aged ≥18 years) with a body-mass index (BMI) of 27 kg/m2 or higher and glycated haemoglobin (HbA1c) of 7–10% (53–86 mmol/mol) were randomly assigned (1:1:1), using a computer-generated random sequence via a validated interactive web-response system, to receive either once-weekly, subcutaneous tirzepatide (10 mg or 15 mg) or placebo for 72 weeks. All participants, investigators, and the sponsor were masked to treatment assignment. Coprimary endpoints were the percent change in bodyweight from baseline and bodyweight reduction of 5% or higher. The treatment-regimen estimand assessed effects regardless of treatment discontinuation or initiation of antihyperglycaemic rescue therapy. Efficacy and safety endpoints were analysed with data from all randomly assigned participants (intention-to-treat population). This trial is registered with ClinicalTrials.gov, NCT04657003.
Between March 29, 2021, and April 10, 2023, of 1514 adults assessed for eligibility, 938 (mean age 54·2 years SD 10·6, 476 51% were female, 710 76% were White, and 561 60% were Hispanic or Latino) were randomly assigned and received at least one dose of tirzepatide 10 mg (n=312), tirzepatide 15 mg (n=311), or placebo (n=315). Baseline mean bodyweight was 100·7 kg (SD 21·1), BMI 36·1 kg/m2 (SD 6·6), and HbA1c 8·02% (SD 0·89; 64·1 mmol/mol SD 9·7). Least-squares mean change in bodyweight at week 72 with tirzepatide 10 mg and 15 mg was –12·8% (SE 0·6) and –14·7% (0·5), respectively, and –3·2% (0·5) with placebo, resulting in estimated treatment differences versus placebo of –9·6% percentage points (95% CI –11·1 to –8·1) with tirzepatide 10 mg and –11·6% percentage points (–13·0 to –10·1) with tirzepatide 15 mg (all p<0·0001). More participants treated with tirzepatide versus placebo met bodyweight reduction thresholds of 5% or higher (79–83% vs 32%). The most frequent adverse events with tirzepatide were gastrointestinal-related, including nausea, diarrhoea, and vomiting and were mostly mild to moderate in severity, with few events leading to treatment discontinuation (<5%). Serious adverse events were reported by 68 (7%) participants overall and two deaths occurred in the tirzepatide 10 mg group, but deaths were not considered to be related to the study treatment by the investigator.
In this 72-week trial in adults living with obesity and type 2 diabetes, once-weekly tirzepatide 10 mg and 15 mg provided substantial and clinically meaningful reduction in bodyweight, with a safety profile that was similar to other incretin-based therapies for weight management.
Eli Lilly and Company.