Contour
closed sourceVetoing for Multi-Agentic Tasks.
"Vetoing" occurs when one AI agent rejects or overrides another AI agent in a collaborative, multi-agent task. This phenomenon exists in activities such as code deployment, writing verification, and any environment where multiple AI agents are collaborating with one another. My study explores the edge cases (the contours) in which vetoing can cause harmful effects in achieving user-defined goals.
- AI Agents
- Multiplayer AI
- Research
- Memory
- Sandboxing
- GPT-5.6
Closed-source project. Details available upon request.