1 paper · 1 filter
Simon Parkinson, Mauro Vallati, Zhipeng Wang +1
When several tools can complete a task, an agent that picks the most powerful one holds more authority than the task needs. Recent work demonstrates that language-model agents do t…