Find and explore trending articles from around the web in a clutter-free reading mode.
theatlantic.com • Technology • World
Leading AI models are potentially cheating on benchmark tests, raising concerns about the accuracy of claims regarding their intelligence and progress.
theverge.com • Technology • World
Meta's release of Llama 4 models, particularly Maverick, raised concerns after it was revealed that the version used to achieve high rankings on LMArena was a customized, experimental version not available to the public.