Multiagent Protocols with Aggregated Confidence Signals

Summary

Notes that no existing method produces or evaluates a confidence for the output of a multi-agent system as a whole - prior work uses confidence only within debate, to weight messages, trigger debate, or calibrate individual agents. It introduces three protocols that first transform raw confidence signals to make them comparable across models and then combine them by soft voting or a probability fusion the authors call Bayesian fusion, yielding aggregated confidence substantially more discriminative than the best single agent or standard debate baselines while correctness stays stable. A concrete, implementable specification for the confidence-weighted arm of the paper's aggregation factor.

Contribution

Queued — not yet read.

Method

Queued — not yet read.

Relevance to RISE

Surfaced by the 2026-08-29 literature sweep as a medium-priority candidate for the Architecture as Epistemology (ISR) paper (angle: aggregating LLM judgments). Verified against: Opened https://arxiv.org/abs/2606.13591 (arXiv abstract page, v1 11 Jun 2026)

Critique / open questions

Queued — not yet assessed.

Key quotes

Confidence is used for reliability, oversight, and a range of downstream decision tasks in Natural Language Processing (NLP), yet no existing method produces or evaluates a confidence for the output of a multiagent system. (abstract)

Prior work uses confidence within multiagent debate (MAD) to weight messages, trigger debate, or calibrate individual agents, but it never aggregates these into a single confidence for the system itself. (abstract)