Canonical preprint (figshare DOI): https://doi.org/10.6084/m9.figshare.30645869
This repository is a public landing page for the preprint.
For citation, always use the DOI above.
We present a Meta-Coordination Layer (MCL) that transfers human organizational practices to multi-agent collaboration, improving output quality on complex, open-ended tasks. MCL encodes workflows as a task graph G=(V, E) with role contracts and deliberation-to-convergence checks. In two comparative studies, MCL was preferred by ~80% (academia, n=14) and ~90% (business, n=45) over strong single-model baselines, with absolute score gains of about +0.5 and +0.9. Collaboration increases cost (5–10x tokens, 2–10x latency), making MCL best for quality-first scenarios.
- PDF (figshare): https://doi.org/10.6084/m9.figshare.30645869
- Project: https://www.TuringCorp.net/tools
- Code/config/scripts (if any): MIT License (see
LICENSE). - Paper text & figures: CC BY 4.0 (see
LICENSE-CC-BY-4.0.md).
Canonical DOI: https://doi.org/10.6084/m9.figshare.30645869
