convolutional-debate-agent

Multi-model orchestration generates diverse solver drafts across multiple models, challenges them with adversarial reviews, and synthesizes a reliability-weighted final answer to reduce single-model bias.

Reliability-weighted aggregation scores and weighs solver drafts from multiple LLMs during a debate pipeline, applying domain-aware weighting to produce a final synthesis with improved trust.

Invoke the debate pipeline by entering the slash command /debate followed by your architecture question, which triggers parallel model generation, adversarial reviews, and final synthesis.

Yes, model selection routing and per-model prompting overrides can be configured in settings to ensure reproducible behavior across multi-model verification and debate sessions.

Use cross-model validation with adversarial reviews for high-stakes or ambiguous reasoning tasks, complex problem solving, and rigorous research synthesis where single-model bias is a concern.

Yes, the multi-model debate pipeline supports optional Aristotle formal verification alongside RWEA scoring to provide formal checks and domain-aware weighting for improved answer trust.