AI
インコンテキスト検索はいつ役立つのか?リフレクション駆動型推論のサンプリング複雑性理論
When Does In-Context Search Help? A Sampling-Complexity Theory of Reflection-Driven Reasoning
arxiv2026年7月9日
日本語要約
本論文は、LLMにおけるリフレクション駆動型推論にインコンテキスト検索がいつ役立つかを説明するための理論的枠組み、すなわちサンプリング複雑性理論を提示する。その重要性は、複雑な推論がどのように現れるかについてのより深く科学的な理解を提供することにある。これは、より有能で予測可能なAI推論システムの設計における根本的に新しいアプローチにつながる可能性がある。
English Summary
This paper presents a theoretical framework, a sampling-complexity theory, to explain when in-context search aids reflection-driven reasoning in LLMs. Its importance lies in providing a deeper, scientific understanding of how complex reasoning emerges. This could lead to fundamentally new approaches in designing more capable and predictable AI reasoning systems.