基于多智能体辩论的创意生成：辩论会抑制多样性吗？
Creative Generation via Multi-Agent Debate: Does Debate Suppress Diversity?
浏览论文内容
中文总结 AI 辅助
针对多智能体辩论（MAD）抑制创意生成多样性的问题，提出Creative-MAD方法，通过认知视角分配和基于嵌入的同伴选择提升多样性并保持输出质量。
中文摘要 AI 辅助
创意生成任务，如叙事写作和科学构思，既需要高质量输出，也需要在独立运行时产生不同的响应以最大化探索性。多智能体辩论（MAD）已在事实和推理任务中展现出强大的质量提升，因此成为创意生成的自然候选方案。然而，我们发现其收敛驱动的设计会主动抑制独立运行时输出的多样性，与创意任务形成固有权衡。我们从理论上证明，在每次辩论会话中保持智能体间的多样性是实现独立运行时多样输出的必要条件。基于该发现，我们提出Creative-MAD，引入两种协同干预措施以维持智能体分歧：具体而言，认知视角分配通过将每个智能体锚定到独特且持久的认知模式来对抗身份漂移，而基于嵌入的同伴选择通过将每个智能体的上下文限制在语义上最远的同伴来对抗多数倾向。在四个创意基准上的实验表明，Creative-MAD在保持MAD输出质量的同时，显著提升了词汇和语义多样性。
英文摘要
Creative generation tasks, such as narrative writing and scientific ideation, demand both high-quality outputs and distinct responses across independent runs to maximize exploration. Multi-Agent Debate (MAD) has shown strong quality gains on factual and reasoning tasks, making it a natural candidate for creative generation. However, we find its convergence-driven design actively suppresses output diversity across independent runs, creating an inherent trade-off with creative tasks. We theoretically show that preserving diversity among agents within each debate session is a necessary condition for achieving diverse outputs across independent runs. Building on this finding, we propose Creative-MAD, which introduces two synergistic interventions to sustain agent divergence. Specifically, Cognitive Lens Assignment counters identity drift by anchoring each agent to a distinct and persistent cognitive mode, while Embedding-based Peer Selection counters majority pull by limiting each agent's context to its most semantically distant peers. Experiments across four creative benchmarks demonstrate that Creative-MAD significantly enhances both lexical and semantic diversity while maintaining MAD's output quality.
发表机构
- Deakin University（迪肯大学）
- Deakin Applied Artificial Intelligence Initiative（迪肯应用人工智能倡议）
机构由 AI 辅助整理，请以论文原文为准。
