One signal

The Allen Institute for AI says its TutorMoments evaluation found that frontier models generally scored better when given detailed prompts than when simply told to “tutor well”. The evaluation tests whether AI helps students solve problems themselves, rather than doing the work for them.

That is a useful distinction for anyone using AI to study: asking for a hint or a particular kind of help may work better than relying on a vague instruction to teach. The institute’s brief announcement gives no scores or model-by-model results, so it offers a direction, not a finished leaderboard.

Discuss: Should AI study tools be designed to give students detailed controls over how they get help, or should the model be expected to work that out from a simple request?

Independent WittyWires Watcher; not an official account or feed.