Audits of multilingual resources are reporting shockingly poor quality: “less than 50% … acceptable quality.” There is too much translationese in too many of our multilingual resources, e.g., Wikipedia, XNLI, FLORES, WordNet. We view translationese as a form of noise that makes it hard to generalize from a benchmark based on translation to a real task of interest that does not involve translation. Worse, too much of this translationese is in the “wrong” direction. Directionality matters. Professional translators translate from their weaker language into their stronger language. Unfortunately, many of our resources translate in the other direction, from a stronger (higher-resource) language into a weaker (lower-resource) language. In Wikipedia, for example, there is more translation out of English than into English. We recommend more investments in high-quality data, and less in translation, especially in the “wrong” direction.

Retrieval-augmented generation (RAG) adds a simple but powerful feature to chatbots, the ability to upload files just-in-time. Chatbots are trained on large quantities of public data. The ability to upload files just-in-time makes it possible to reduce hallucinations by filling in gaps in the knowledge base that go beyond the public training data such as private data and recent events. For example, in a customer service scenario, with RAG, we can upload your private bill and then the bot can discuss questions about your bill as opposed to generic FAQ questions about bills in general. This tutorial will show how to upload files and generate responses to prompts; see https://github.com/kwchurch/RAG for multiple solutions based on tools from OpenAI, LangChain, HuggingFace transformers and VecML.

Large language models (LLMs) have achieved amazing successes. They have done well on standardized tests in medicine and the law. That said, the bar has been raised so high that it could take decades to make good on expectations. To buy time for this long-term research program, the field needs to identify some good short-term applications for smooth-talking machines that are more fluent than trustworthy.

The Visean–Serpukhovian boundary is not yet defined by a Global Stratotype Section and Point (GSSP) but it is recognizable operationally by the appearance of the conodont Lochriea ziegleri in the L. nodosa–L. ziegleri chronocline. Foraminiferal successions across this boundary in the type area of the Serpukhovian Stage (Moscow Basin, Russia), elsewhere in Russia and in the central United States suggest that the appearances of Asteroarchaediscus postrugosus, Janischewskina delicata, Eolasiodiscus donbassicus, and specimens controversially referred to “Millerella tortula” are reliable, auxiliary indices to the base of the Serpukhovian. In southern Guizhou Province, China, Visean–Serpukhovian rock sequences from slope and platform settings have yielded rich associations of conodonts and foraminifers, respectively. The Nashui section is a leading candidate for the Serpukhovian GSSP because its slope deposits contain an uninterrupted record of conodont occurrences including the L. nodosa–L. ziegleri transition. Foraminifers recovered from the Nashui section are comparatively rare and include none of the basal Serpukhovian indices. In contrast, the nearby Yashui section represents a platform interior setting in which foraminifers flourished and conodonts were nearly absent. The base of the Serpukhovian at Yashui is marked approximately by the appearance of “tortula-like” specimens. Although it is not possible to correlate biostratigraphically between the Nashui and Yashui sections, the occurrence of “tortula-like” specimens at the Yashui section allows correlation with the mid-Venevian Substage of the Moscow Basin at a level coinciding with the appearance of L. ziegleri. Together, the slope and platform sections comprise an informative biostratigraphic reference area for micropaleontologic characterization of the Visean–Serpukhovian boundary in southern Guizhou.

The NASA Kepler satellite has provided unprecedented high duty-cycle, high-precision light curves for a large number of stars by continuously monitoring a field of view in Cygnus-Lyra region, leading to great progress in both discovering exoplanets and characterizing planet-hosting stars by means of asteroseismic methods. The asteroseismic survey allows the investigation of stars covering the whole H-R diagram. However, the low precision of effective temperatures and surface gravities in the KIC catalogue and the lack of information on chemical composition, metallicity and rotation rate prevent asteroseismic modeling, requiring spectroscopic observations for thousands of asteroseismic targets in the Kepler field in a homogeneous way.

Short-chain oligomers of aniline are attractive semi-metallic materials for applications as organic electrodes or hole-transporting layers in organic photovoltaics. However, conventionally processed oligoanilines are often amorphous, which limits their conductivities and carrier transport mobilities. Here, we report a simple solvent-exchange method that can render a variety of oligoanilines and their derivatives into crystals of different shapes and dimensions, including 1-D fibers and wires, 2-D ribbons, and 3-D plates, hollow spheres, porous sheets, and flower-like structures. Dopant ions are also simultaneously incorporated into the crystals during self-assembly, allowing them to become conducting. Mechanistic studies suggest that the higher order crystals arise from the most primitive nanofibrillar morphology via hierarchical assembly, providing insights into a general approach to control organic crystal morphologies. Selected area electron diffraction studies reveal their single crystalline nature.

We present results of extremely short carrier lifetime measurements on a series of hydrogenated nanocrystalline silicon (nc-Si:H) thin films by a novel, non-destructive, non-contact method. Transmission modulated photoconductive decay (TMPCD) is a newly developed technique which appears to have high enough sensitivity and time resolution to measure the extremely short carrier lifetimes on the order of a nanosecond. As a proof of this, we measure various nc-Si:H samples of varying crystalline volume fraction as well as a fully amorphous sample. To ascribe an effective lifetime to the materials, we use a simple model which assumes a single exponential decay. By using this model, effective lifetimes can be deconvoluted from our pump beam giving nanosecond lifetimes. Lifetimes of between 1.9 and 0.9 nanoseconds are reported and trend to decreasing lifetimes as crystalline volume fraction is increased.

Mechanical milling provides a unique means of studying the influence of grain size and disorder on the magnetic properties of nanostructured alloys. This paper compares the role of milling in the nanostructure evolution of two ferromagnets – SmCo5 and GdAl2 – and the subsequent impact of nanostructure on magnetic properties and phase transitions. The ferromagnetic properties of SmCo5 are enhanced by short (< 2 hours) milling times, producing up to an eight-fold increase in coercivity and high remanence ratios. The coercivity increase is attributed to defect formation and strain. Additional milling increases the disorder and produces a mix of ferromagnetic and antiferromagnetic interactions that form a magnetically glassy phase. GdAl2, which changes from ferromagnetic in its crystalline form to spin-glass-like in its amorphous form, is a model system for studying the dependence of magnetically glassy behavior on grain size and disorder. Nanostructured GdAl2 with a mean grain size of 8 nm shows a combination of ferromagnetic and magnetically glassy behavior, in contrast to previous studies of nanostructured GdAl2 with a grain size of 20 nm that show only spin-glass-like behavior.