Search papers, labs, and topics across Lattice.
2
0
3
0
This work audits five LLM overseers on 19 compliance tasks, varying only trace detail and evidence labeling, finding that elaborate traces shift the decision criterion toward rejection, increasing false alarms in susceptible overseers.
LVLMs can coordinate efficient communication when explicitly prompted, but struggle to do so with implicit cues, revealing a critical gap in AI-human communication parallels.
