Search for: All records

Large language models (LLMs) demonstrate impressive reasoning abilities, but translating reasoning into actions in the real world remains challenging. In particular, it is unclear how to complete a given task provably within a minimum number of interactions with the external environment, e.g., through an internal mechanism of reasoning. To this end, we propose the first framework with provable regret guarantees to orchestrate reasoning and acting, which we call “reason for future, act for now” (RAFA). Specifically, we design a prompt template for reasoning that learns from the memory buffer and plans a future trajectory over a long horizon (“reason for future”). At each step, the LLM agent takes the initial action of the planned trajectory (“act for now”), stores the collected feedback in the memory buffer, and reinvokes the reasoning routine to replan the future trajectory from the new state. The key idea is to cast reasoning in LLMs as learning and planning in Bayesian adaptive Markov decision processes (MDPs). Correspondingly, we prompt LLMs with the memory buffer to estimate the unknown environment (learning) and generate an optimal trajectory for multiple future steps that maximize a value function (planning). The learning and planning subroutines are performed in an “incontext” manner to emulate the actor-critic update for MDPs. Our theoretical analysis establishes a √T regret, while our experimental validation demonstrates superior empirical performance. Here, T denotes the number of online interactions. 

ABSTRACT The circular-crack model has been widely used in seismology to infer earthquake stress drop. A common assumption is that the background medium is isotropic, although many earthquakes occur in geologically anisotropic settings. In this article, we study the effect of anisotropy on stress drop for a circular crack model and present explicit formalism in both static and kinematic cases. In the static case, we obtain the relationship between stress drop and slip for a circular crack model in an arbitrarily anisotropic medium. Special attention is given to the transversely isotropic (TI) medium. The static formalism is useful in understanding stress drop, but not all quantities are observables. Therefore, we resort to the kinematic case, from which we can infer stress drop using recorded far-field body waves. In the kinematic case, we assume that the crack ruptures circularly and reaches the final displacement determined by the static solutions. The far-field waveforms show that the corner frequency will change with different anisotropic parameters. Finally, we calculate the stress drops for cracks in isotropic and anisotropic media using the far-field waveforms. We find that in an isotropic medium, only shear stress acting on the crack surface contributes to shear slip. However, in a TI medium, if the anisotropy symmetry axis is not perpendicular or parallel to the crack surface, a normal stress (normal to the crack surface) can produce a shear slip. In calculating stress drop for an earthquake in an anisotropic medium using far-field body waves, a large error may be introduced if we ignore the possible anisotropy in the inversion. For a TI medium with about 18% anisotropy, the misfit of inferred stress drop could be up to 41%. Considering the anisotropic information, we can further improve the accuracy of stress-drop inversion. 

BackgroundDespite significant global progress in reducing neonatal mortality, bacterial sepsis remains a major cause of neonatal deaths.Klebsiella pneumoniae(K.pneumoniae) is the leading pathogen globally underlying cases of neonatal sepsis and is frequently resistant to antibiotic treatment regimens recommended by the World Health Organization (WHO), including first-line therapy with ampicillin and gentamicin, second-line therapy with amikacin and ceftazidime, and meropenem. Maternal vaccination to prevent neonatal infection could reduce the burden ofK.pneumoniaeneonatal sepsis in low- and middle-income countries (LMICs), but the potential impact of vaccination remains poorly quantified. We estimated the potential impact of such vaccination on cases and deaths ofK.pneumoniaeneonatal sepsis and project the global effects of routine immunization of pregnant women with theK.pneumoniaevaccine as antimicrobial resistance (AMR) increases. Methods and findingsWe developed a Bayesian mixture-modeling framework to estimate the effects of a hypotheticalK.pneumoniaematernal vaccine with 70% efficacy administered with coverage equivalent to that of the maternal tetanus vaccine on neonatal sepsis infections and mortality. To parameterize our model, we used data from 3 global studies of neonatal sepsis and/or mortality—with 2,330 neonates who died with sepsis surveilled from 2016 to 2020 undertaken in 18 mainly LMICs across all WHO regions (Ethiopia, Kenya, Mali, Mozambique, Nigeria, Rwanda, Sierra Leone, South Africa, Uganda, Brazil, Italy, Greece, Pakistan, Bangladesh, India, Thailand, China, and Vietnam). Within these studies, 26.95% of fatal neonatal sepsis cases were culture-positive forK.pneumoniae. We analyzed 9,070K.pneumoniaegenomes from human isolates gathered globally from 2001 to 2020 to quantify the temporal rate of acquisition of AMR genes inK.pneumoniaeisolates to predict the future number of drug-resistant cases and deaths that could be averted by vaccination.Resistance rates to carbapenems are increasing most rapidly and 22.43% [95th percentile Bayesian credible interval (CrI): 5.24 to 41.42] of neonatal sepsis deaths are caused by meropenem-resistantK.pneumoniae. Globally, we estimate that maternal vaccination could avert 80,258 [CrI: 18,084 to 189,040] neonatal deaths and 399,015 [CrI: 334,523 to 485,442] neonatal sepsis cases yearly worldwide, accounting for more than 3.40% [CrI: 0.75 to 8.01] of all neonatal deaths. The largest relative benefits are in Africa (Sierra Leone, Mali, Niger) and South-East Asia (Bangladesh) where vaccination could avert over 6% of all neonatal deaths. Nevertheless, our modeling only considers country-level trends inK.pneumoniaeneonatal sepsis deaths and is unable to consider within-country variability in bacterial prevalence that may impact the projected burden of sepsis. ConclusionsAK.pneumoniaematernal vaccine could have widespread, sustained global benefits as AMR inK.pneumoniaecontinues to increase. 

Paysonia auriculata (Brassicaceae) produces multiple hydroxy fatty acids as major components of the seed oil. We tracked the changes in seed oil composition and gene expression during development, starting 14 days after flowers had been pollinated. Seed oil changes showed initially higher levels of saturated and unsaturated fatty acids (FAs) but little accumulation of hydroxy fatty acids (HFAs). Starting 21 days after pollination (DAP) HFA content sharply increased, and reached almost 30% at 28 DAP. Total seed oil also increased from a low of approximately 2% at 14 DAP to a high of approximately 20% by 42 DAP. We identified almost all of the fatty acid synthesis and modification genes that are known from Arabidopsis, and, in addition, a strong candidate for the hydroxylase gene that mediates the hydroxylation of fatty acids to produce valuable hydroxy fatty acids (HFAs) in this species. The gene expression network revealed is very similar to that of the emerging oil crop, Physaria fendleri , in the sister genus to Paysonia . Phylogenetic analyses indicate the hydroxylase enzyme, FAH12, evolved only once in Paysonia and Physaria , and that the enzyme is closely related to FAD2 enzymes. Phylogenetic analyses of FAD2 and FAH12 in the Brassicaceae and outgroup genera suggest that the branch leading to the hydroxylase clade of Paysonia and Physaria is under relaxed selection, compared with the strong purifying selection found across the FAD2 lineages. 

Key Points The 15 January 2022 Hunga Tonga‐Hunga Ha'apai eruption had four episodic seismic subevents with similar waveforms within ∼300 s An unusual upward force jump‐started each subevent A magma hammer explains the force and estimates the subsurface magma mass flux which fits the vent discharge rate based on satellite data 

Abstract Abscission, known as shattering in crop species, is a highly regulated process by which plants shed parts. Although shattering has been studied extensively in cereals and a number of regulatory genes have been identified, much diversity in the process remains to be discovered. Teff (Eragrostis tef) is a crop native to Ethiopia that is potentially highly valuable worldwide for its nutritious grain and drought tolerance. Previous work has suggested that grain shattering in Eragrostis might have little in common with other cereals. In this study, we characterize the anatomy, cellular structure, and gene regulatory control of the abscission zone (AZ) in E. tef. We show that the AZ of E. tef is a narrow stalk below the caryopsis, which is common in Eragrostis species. X-ray microscopy, scanning electron microscopy, transmission electron microscopy, and immunolocalization of cell wall components showed that the AZ cells are thin walled and break open along with programmed cell death (PCD) at seed maturity, rather than separating between cells as in other studied species. Knockout of YABBY2/SHATTERING1, documented to control abscission in several cereals, had no effect on abscission or AZ structure in E. tef. RNA sequencing analysis showed that genes related to PCD and cell wall modification are enriched in the AZ at the early seed maturity stage. These data show that E. tef drops its seeds using a unique mechanism. Our results provide the groundwork for understanding grain shattering in Eragrostis and further improvement of shattering in E. tef.