Search papers, labs, and topics across Lattice.
Affiliation:
4
0
6
25
Hallucinated vulnerabilities from LLMs are not just a nuisance; they create a cognitive burden that can overwhelm cybersecurity triage systems.
Reliable detection of malicious agent skills hinges on a comprehensive benchmark that reveals stark differences in threat composition across sources.
Auditing Chinese web content reveals pervasive pollution that shifts over time, challenging the integrity of LLM training data.
Prompt complexity is a critical dimension that significantly influences maintenance effort, challenging traditional views that prioritize code-level metrics alone.
