Training on the evalA repeatable test for AI quality — a set of tasks plus scoring — used the way software teams use test suites, because model output is too variable to judge by eyeballing.Full definition → distribution cut blackmail rates without improving held-out alignmentThe work of making AI systems actually pursue what their builders and users intend, rather than something subtly or dangerously different.Full definition →. Teaching the reasons behind good behavior generalized better, so teams fine-tuningTaking a trained model and training it a bit more on your own examples so it gets better at one specific job.Full definition → agents should not rely on demonstrations alone.
Terms in this piece · Glossary
eval — A repeatable test for AI quality — a set of tasks plus scoring — used the way software teams use test suites, because model output is too variable to judge by eyeballing.
alignment — The work of making AI systems actually pursue what their builders and users intend, rather than something subtly or dangerously different.
fine-tuning — Taking a trained model and training it a bit more on your own examples so it gets better at one specific job.
