Showing cs.AIShow all
2 papers · 1 filter
cs.AI2026
LACUNA: Safe Agents as Recursive Program Holes
Yaoyu Zhao, Yichen Xu, Oliver BraÄevac +3
LLM agents increasingly act by writing code, yet a split persists between the runtime that drives the agent and the code the model writes. The runtime owns the loop, context, and c…
cs.AI2026
Tracking Capabilities for Safer Agents
Martin Odersky, Yaoyu Zhao, Yichen Xu +2
AI agents that interact with the real world through tool calls pose fundamental safety challenges: agents might leak private information, cause unintended side effects, or be manip…