Topic
How a single institutional announcement frames a novel interpretive claim, including hedging and absence of external comment.
OpenAI's louder announcement of a nine-member independent math advisory group caused widespread confusion over whether the company had appointed it. Its first job is helping release scores more results from an unreleased OpenAI model, a wave researchers already dread.
Reality
Ramsauer and colleagues proved the equality in 2020, three years after attention was published. It holds for a single retrieval step inside a forward pass, and the capacity figures quoted alongside it come from a different model.
Reality
A Binghamton team scored each candidate word by how much uncertainty it was expected to remove and reported solving 99 percent of simulated Wordle puzzles. The university did not say how many games it simulated.
Reality
Quanta put a simple question to the mathematicians extending what has been billed as a grand unified theory of mathematics. The cause of their disagreement is structural. A sweeping claim rests on very technical individual results, in a discipline that declines to interpret what it cannot prove.
Reality
Australian coverage of sexual violence against women over 60 favours stranger attacks and visible injuries, while the aged care count runs near 50 assaults a week, most by partners or acquaintances. Researchers say credibility follows the frail stereotype.
Reality
A Southampton-led team models 100 million years of East Antarctic uplift and puts the ice sheet's origin in the deep Earth, not the atmosphere.
Reality
A Penn State team pulled native RNA polymerase II complexes out of fruit fly embryos and found some missing two of the canonical 12 subunits. The purified picture was tidier than the cell.
Reality
A Tel Aviv University team reports that male Neanderthal pelvises track modern human females, not males, inverting a baseline used for decades to read hominin hips.
Reality