This toot from Charlie Gao about the new release of {mirai} (R’s newest async evaluation framework) mentions the new function race_mirai() which > returns as soon as the first mirai you pass to it resolves. This supports more async usage patterns in #RStats.This caught my attention, partly because I’ve been thinking about “alternate pathway” designs recently, albeit in a very different context. I saw the latest paper from Thinking Machines Lab titled ‘Defeating Nondeterminism in LLM Inference’ and the header “image” (actually a complex svg, not a rasterised image) for that shows the result of an LLM annotated with alternatives which were not selected (but might have been with higher temperature) and their ‘score’ for how likely they were to be selected (presumably)This reminded me of Tufte’s graphical sentences, apparently in his book ‘Seeing with Fresh Eyes’ but which I could only find one blog post referencing.

Use the constructive package instead of dput() to construct R objects with idiomatic R code for human-readable reproducible examples

This book provides an overview of skills needed for reproducible and open research using the statistical programming language R and tidyverse packages. It covers reproducible workflows, data visualisa...

Welcome to the inaugural Charlottesville R Users (CRU) meetup! In this first event we'll discuss the format and goals of CRU, followed by a short technical…