Mamba dropped attention to make long sequences cheap. Its authors put some back
Mamba matches a transformer twice its size on short-context reasoning up to 2.8B parameters, and loses to a far smaller one at looking things up in a long context.
Whisper answers silence with a sentence no one spoke
The stage that writes the words is a language model, and a language model with almost nothing to go on still produces its most probable sentence instead of a blank.
How hard the testers push decides what a safety test finds
The same model scored 17.9% and then 56.9% on the same test after nothing but its prompt changed, so no single safety score is the ceiling on what a system can do.
Google's eye-screening AI rejected a fifth of the photos nurses took in Thai clinics
The model diagnosed diabetic eye disease as accurately as a retina specialist. Its quality bar, tuned high for patient safety, collided with undarkened rooms, slow clinic internet, and photographs it judged too flawed to grade.
BERTScore grades how much two sentences look alike to a language model
Introduced in 2020 to go past word-counting scores, it rates a candidate by how closely a pretrained encoder sees it resembling the reference, which rewards fluent wrong answers and carries the encoder's blind spots.