BraveGuard trains guard models on realistic agent trajectories derived from open-world threats, raising detection accuracy on AgentHazard from 38.79% to 82.38%.
Guardtrace-vl: Detecting unsafe multimodel reasoning via iterative safety supervision
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CR 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
BraveGuard: From Open-World Threats to Safer Computer-Use Agents
BraveGuard trains guard models on realistic agent trajectories derived from open-world threats, raising detection accuracy on AgentHazard from 38.79% to 82.38%.