flâneur

Can Frontier Models Autocomplete Safety Research? — LessWrong

lesswrong.com · saved by 1 readers

We pose the following research question: how can we measure the "research taste" of language models in experiment planning? What parts of planning ta…

saved by