The process begins with feeding an algorithm enormous amounts of data—books, math problems, captioned photos, voice recordings, and so on—to establish the model’s baseline capabilities.

Switzerland has just released Apertus, its open-source national Large Language Model (LLM) that it hopes would be an ...

Underspecification means something different: even if a training process can produce a good model, it could still spit out a bad one because it won’t know the difference. Neither would we.

“There has been this long-hypothesized failure mode, which is that you'll run your training process, and all the outputs will look good to you, but the model is plotting against you,” says ...

In July, EPFL, ETH Zurich, and CSCS announced their joint initiative to build a large language model (LLM). Now, this model ...