[论文] Consensus is Strategically Insufficient: Reasoning-Trace Disagreement ...

论文概要

研究领域: ML
作者: Michał Wawer, Jarosław A. Chudziak
发布时间: 2025-06-01
arXiv: 2506.00633

中文摘要

多智能体系统通常通过投票、共识协议、辩论或容错聚合来减少分歧。本文认为，对于涉及价值判断的任务，这一目标并不充分——因为分歧可能反映真实的规范性不确定性，而非智能体错误。基于先前关于人机协作审核中推理痕迹分歧的工作，我们提出一种知识表征层，将推理痕迹和智能体决策抽象为符号化的分歧状态。对于产生显式推理痕迹和二值决策的智能体，我们根据推理相似性和结论一致性区分四种状态：收敛一致、发散一致、收敛分歧、发散分歧。这些状态支持可废止的战略路由规则。我们在内容审核中实例化该框架，并论证分歧感知路由为子符号LLM推理与符号知识表征之间的多智能体战略推理提供了桥梁。

原文摘要

Multi-agent systems are commonly designed to reduce disagreement through voting, consensus protocols, debate, or fault-tolerant aggregation. We argue that this objective is insufficient for value-laden tasks, where disagreement may reflect genuine normative uncertainty rather than agent error. Building on prior work on reasoning-trace disagreement in human-AI collaborative moderation, we propose a knowledge-representation layer in which reasoning traces and agent decisions are abstracted into symbolic disagreement states. Given agents producing explicit reasoning traces and binary decisions, we distinguish four states according to reasoning similarity and conclusion agreement: convergent agreement, divergent agreement, convergent disagreement and divergent disagreement. These states suppor...

自动采集于 2026-06-05

#论文 #arXiv #ML #小凯