Fully-Connected Debate
N agents in a fully-connected mesh iterate arguments and counter-arguments each round until consensus. Every agent sees every other's proposal per round.
Verified against arxiv.org
In a debate topology, N agents in a fully-connected mesh argue a question: each sees the others' positions, rebuts and revises across rounds, until they converge or a judge decides. The friction is the point: adversarial cross-examination surfaces errors a single agent would miss. It suits open questions where reasoning quality matters more than speed. In anatomy terms it is N reasoning cores sharing a common transcript, iterating.
N reasoning cores are fully connected over a shared transcript; rounds of argue, rebut and revise run until convergence; an optional judge core resolves; effectors act only on the settled answer.
- ·Open, contestable questions in strategy, analysis or research.
- ·Red-teaming and error-finding.
- ·Decisions where surfacing counter-arguments matters more than speed.
- ·Surfaces errors and blind spots through adversarial pressure.
- ·Higher-quality reasoning on open questions.
- ·Diverse perspectives by construction.
- ·Self-documenting via the argument transcript.
- ·Expensive: N agents multiplied by rounds.
- ·Can fail to converge or can oscillate.
- ·Verbose: signal gets buried in argument.
- ·Groupthink if the agents are too similar.
Description, use cases and trade-offs: Field Book analysis, synthesised from established agent-architecture literature.
The corpus does not rewrite vendor documentation. Where copy is quoted verbatim it is marked as such and attributed to its source.