Search papers, labs, and topics across Lattice.
Affiliation:
3
0
5
1
Trajectory-scoped safety monitors for LLM agents are fundamentally flawed, leading to a 50% chance of missing threats that span multiple iterations.
Social bot detection gets a boost from anisotropic hyperbolic geometry, which adapts to varying structural densities and outperforms existing methods by decoupling account features from potentially contaminated neighborhood aggregations.
LLM agent harnesses are surprisingly vulnerable, but weaving security directly into the agent lifecycle can slash attack success by 42% without sacrificing utility.
