---
id: 20260917-T0-06
title: "State of Thought：让LLM靠内部状态自主推理，减少外部控制依赖"
title_en: "State of Thought Enables LLMs to Reason Endogenously Without External Control"
url: https://ai.daily.yangsir.net/daily/20260917-T0-06
issue_date: 2026-09-17
publish_date: 2026-09-16T04:00:00.000Z
category: research
source_name: "arXiv cs.CL (NLP)"
source_url: https://arxiv.org/abs/2609.16055
---

# State of Thought：让LLM靠内部状态自主推理，减少外部控制依赖

新论文提出State of Thought方法，让大语言模型通过内部状态实现自主推理，不再依赖固定推理程序等外部控制机制。现有测试时推理方案大多依靠人为设定的推理步骤或外部调度，限制了模型的灵活性和泛化能力。该方法将推理能力内化到模型自身状态中，在测试时计算框架下探索更自主的推理路径。如果效果得到验证，这意味着开发者可以减少对提示工程和外部控制逻辑的依赖，让模型在复杂任务中自行决定推理深度和方式。

## English Version

**State of Thought Enables LLMs to Reason Endogenously Without External Control**

A new paper introduces State of Thought, a method that enables LLMs to perform endogenous reasoning through internal states rather than relying on externally imposed control such as fixed reasoning programs. Existing test-time reasoning paradigms depend heavily on human-designed reasoning steps, which limits flexibility and generalization. This approach internalizes reasoning into the model's own state, exploring more autonomous reasoning paths under test-time compute. If validated, developers could reduce dependence on prompt engineering and external control logic, letting models decide reasoning depth and strategy on their own.

---

**来源**：[arXiv cs.CL (NLP)](https://arxiv.org/abs/2609.16055)

**详情页**：https://ai.daily.yangsir.net/daily/20260917-T0-06

---

*智语观潮 · Daily — https://ai.daily.yangsir.net/llms.txt*