Collaboration-of-Experts (CoE)
Updated 2026-08-19
INTRODUCTION
English translation pending.
CORE DEFINITION
In AI system design, a collaboration-of-experts pipeline assigns distinct roles to separate models: one produces an initial answer, another reviews it for errors and omissions, and a third integrates the corrections into a final output. The loop can repeat until the answer stabilizes. The design mirrors the human decision chain of generation, criticism and synthesis, and it trades additional compute for greater accuracy on complex problems where a single pass is unreliable.
SCAFFOLDING EFFECT
Reduce cognitive load
- Split the roles: separate generation, criticism and synthesis instead of doing all three at once. - Force the critique: require an explicit review stage before accepting any draft. - Iterate to stability: repeat the loop until further rounds stop changing the answer.
Anchor fast decisions
A single pass produces an answer without any internal check, so errors pass through unnoticed. Separating the critic into a distinct step creates an explicit opportunity to catch them, and integration applies the corrections rather than discarding the draft. Each additional round reduces the error rate, though with diminishing returns and rising cost.
MINIMUM ACTION
In progress 0/1Practice this model in one real situation:
account_treeGenealogyexpand_more
menu_bookReferencesexpand_more
Source support: Explicit
- arxiv.orghttps://arxiv.org/abs/2107.03815verified
PRIVATE NOTES · Only visible to you
SAVED Q&A
ENTRY Q&A · Private saving available
Ask with a clear boundary
thinkingmodels answers from published entry context only.
Your question is sent to thinkingmodels. The answer uses public entry context only.
RELATED MODELS