Jove
Visualize
联系我们
JoVE
x logofacebook logolinkedin logoyoutube logo
关于 JoVE
概览领导团队博客JoVE 帮助中心
作者
出版流程编辑委员会范围与政策同行评审常见问题投稿
图书馆员
用户评价订阅访问资源图书馆顾问委员会常见问题
研究
JoVE JournalMethods CollectionsJoVE Encyclopedia of Experiments存档
教育
JoVE CoreJoVE BusinessJoVE Science EducationJoVE Lab Manual教师资源中心教师网站
使用条款与条件
隐私政策
政策

相关概念视频

Schemas01:42

Schemas

11.5K
A schema is a mental construct consisting of a cluster or collection of related concepts (Bartlett, 1932). There are many different types of schemata, and they all have one thing in common: schemata are a method of organizing information that allows the brain to work more efficiently. When a schema is activated, the brain makes immediate assumptions about the person or object being observed.
11.5K

您也可能阅读

相关文章

通过共同作者、期刊和引用图与本文相关的文章。

排序
Same author

LazyAct: Lazy actor with dynamic state skip based on constrained MDP.

PloS one·2025
Same author

Single-Atom Rh on High-Index CeO<sub>2</sub> Facet for Highly Enhanced Catalytic CO Oxidation.

Angewandte Chemie (International ed. in English)·2023
Same author

New Strategy: Molten Salt-Assisted Synthesis to Enhance Lanthanide Upconversion Luminescence.

Small (Weinheim an der Bergstrasse, Germany)·2023
Same author

Facile synthesis of Sb<sup>3+</sup>-doped (Bmim)<sub>2</sub>InCl<sub>5</sub>(H<sub>2</sub>O) through a grinding method for light-emitting diodes.

Dalton transactions (Cambridge, England : 2003)·2023
Same author

Boosting Polyethylene Hydrogenolysis Performance of Ru-CeO<sub>2</sub> Catalysts by Finely Regulating the Ru Sizes.

Small (Weinheim an der Bergstrasse, Germany)·2023
Same author

Multiple site occupancy induced yellow-orange emission in an Eu<sup>2+</sup>-doped KSr<sub>6</sub>Sc(SiO<sub>4</sub>)<sub>4</sub> phosphor towards optical temperature sensors.

Dalton transactions (Cambridge, England : 2003)·2023

相关实验视频

Updated: May 22, 2025

Modeling the Functional Network for Spatial Navigation in the Human Brain
05:55

Modeling the Functional Network for Spatial Navigation in the Human Brain

Published on: October 13, 2023

967

通过快速工程来规划路径的大型语言模型中减轻空间幻觉.

Hongjie Zhang1, Hourui Deng2, Jie Ou3

  • 1College of Computer Science, Sichuan Normal University, Chengdu, 610101, China. zhanghongjie@sicnu.edu.cn.

Scientific reports
|March 15, 2025
PubMed
概括

我们介绍了S2ERS,这是一种新的技术,用于改进体现智能的大型语言模型 (LLM) 的空间推理. S2ERS显著减少了幻觉问题,提高了路径规划成功率.

更多相关视频

Augmenting Large Language Models via Vector Embeddings to Improve Domain-Specific Responsiveness
03:14

Augmenting Large Language Models via Vector Embeddings to Improve Domain-Specific Responsiveness

Published on: December 6, 2024

474
The Spatial Memory Game: Testing the Relationship Between Spatial Language, Object Knowledge, and Spatial Cognition
05:15

The Spatial Memory Game: Testing the Relationship Between Spatial Language, Object Knowledge, and Spatial Cognition

Published on: February 19, 2018

10.7K

相关实验视频

Last Updated: May 22, 2025

Modeling the Functional Network for Spatial Navigation in the Human Brain
05:55

Modeling the Functional Network for Spatial Navigation in the Human Brain

Published on: October 13, 2023

967
Augmenting Large Language Models via Vector Embeddings to Improve Domain-Specific Responsiveness
03:14

Augmenting Large Language Models via Vector Embeddings to Improve Domain-Specific Responsiveness

Published on: December 6, 2024

474
The Spatial Memory Game: Testing the Relationship Between Spatial Language, Object Knowledge, and Spatial Cognition
05:15

The Spatial Memory Game: Testing the Relationship Between Spatial Language, Object Knowledge, and Spatial Cognition

Published on: February 19, 2018

10.7K

科学领域:

  • 人工智能的人工智能
  • 自然语言处理自然语言处理.
  • 机器人技术 机器人技术 机器人技术

背景情况:

  • 大型语言模型 (LLM) 是体内智能的基础.
  • 由于幻觉,LLM在迷宫般的环境中扎于空间推理和路径规划.
  • 像思维链 (CoT) 这样的现有方法在解决这些空间幻觉问题方面存在局限性.

研究的目的:

  • 开发一种基于LLM的技术,S2ERS,用于空间推理任务的最佳路径规划.
  • 为了减轻LLMs的空间幻觉问题.
  • 提高LLM在体内情报应用中的性能.

主要方法:

  • S2ERS将实体和关系提取与Sarsa强化学习算法集成在一起.
  • 通过从基于文本的迷宫描述中提取图形结构来解决空间幻觉.
  • 提示工程包括状态动作值函数Q和动态本地Q表,以指导规划和减少代币消费.

主要成果:

  • 在LLMs中,S2ERS显著减轻了空间幻觉.
  • 对ChatGPT 3.5,ERNIE-Bot 4.0和ChatGLM-6B进行的实验评估显示了显著的改进.
  • 与最先进的 (SOTA) CoT 方法相比,成功率提高了大约29%,最佳率提高了19%.

结论:

  • S2ERS为改善LLM空间推理和路径规划提供了有效的解决方案.
  • 该技术提高了LLM在体内情报任务中的可靠性.
  • 对于空间推理挑战,S2ERS比现有的CoT方法有了显著的进步.