computer-using agent

safety threat (1) trajectory annotation (1) ui trajectory (1) security threat (1) long-horizon task (1) large language model (1) in-context exemplar (1) memory-augmented agent (1) multimodal input (1) vision-language model (1)