AI.info

Roman Vainshtein

Explore Roman Vainshtein on AI.info.

• Training-Free Policy Violation Detection via Activation-Space Whitening in LLMs
• Counterfactual-based Agent Influence Ranker for Agentic AI Workflows