179 questions
No questions match those filters.
← All questions
AgentsMedium· Multi-Agent Systems
What are the common communication patterns in multi-age...
This is one of the questions in the full AI/ML interview bank. Pro unlocks all 1789 questions; Premium includes the same bank plus the highest daily Practice limit.
See plansFour shapes cover most multi-agent workflows, and picking the wrong one for the task is a common design mistake:
- Sequential / pipeline — Agent A’s output becomes Agent B’s input, then C’s. An assembly line: researcher, then writer, then editor, in a fixed order. Simple and predictable, but rigid.
- Hierarchical — a manager or orchestrator agent delegates sub- tasks to workers and combines what comes back. Good when the right allocation of work isn’t known in advance and needs a decision-maker in the loop.
- Peer-to-peer / collaborative — agents talk to each other directly, debating or building on shared work rather than going through a manager. Fits open-ended tasks like reaching consensus on a strategy where no single agent should have the final say alone.
- Broadcast — one agent publishes information and every other agent receives it simultaneously, useful when many agents need the same update at once, like a price feed reaching every trading agent watching it.
The choice should follow the task’s actual shape: sequential for a fixed, predictable workflow; hierarchical for complex orchestration where allocation is dynamic; peer-to-peer for creative or debate-style tasks; broadcast for one-to-many state updates. Naming which pattern fits a given scenario, and why, matters more than knowing all four exist.