首页 > AI前沿 > NeoResearch: Literature- and Experience-Guided Autonomous Time Series Model Discovery

NeoResearch: Literature- and Experience-Guided Autonomous Time Series Model Discovery

arXiv机器学习 2026-11-13 22:30 8 阅读 查看原文
用户现在需要我把给定的新闻正文排版成语义化HTML片段,遵循规则。首先看正文,首先整体是一段介绍,然后分部分?不对,正文里的段落,首先开头的段落,然后看内容,要保留所有信息,不增删改。首先,原文的段落,第一部分是开头的内容,然后接下来是具体的NeoResearch的内容,还有实验部分。 首先,先处理每个自然段落,用

包裹。然后看有没有适合分小节的?不对,原文的结构:首先开头讲背景,然后提出NeoResearch,然后具体讲它的工作流程,知识挖掘,记忆机制,实验结果。那要不要用

或

?看规则,长文用

/

划分小节。那首先,比如可以分几个小节?比如“研究背景与框架提出”作为

?不对,要保留原文内容,不能加新的标题文字,哦不对,规则是保留原文所有信息,不得增删改,所以标题的内容必须是原文里的?不对,等下,原文里的结构,是不是不需要额外加标题?不对,看原文的内容,有没有可以划分的小节?等下看原文: 原文内容: 第一部分:Developing effective...to discover lightweight, domain-specific forecasting models. 这部分是背景和提出框架。 然后下一句:Specifically, NeoResearch establishes...through experimental feedback. 讲工作流程。 然后:To enable knowledge-grounded discovery...to formulate task-grounded, testable architecture hypotheses. 讲知识挖掘。 然后:Furthermore, a hierarchical experience memory...supporting reuse across forecasting tasks. 讲记忆机制。 然后:Experiments on ten datasets...Our code is available. 讲实验和代码。 那按照规则,要不要用

或

?如果用的话,标题的文本必须是原文里的吗?不对,规则是保留原文所有信息,不得增删改,所以不能添加原文没有的标题文字?哦对,不能增删改内容,所以不能自己加标题的文本,那是不是如果没有明确的小节标题,就用

包裹每个自然段?不对,看原文的自然段,原文是一整段?不,用户给的正文是一段?不对,看用户给的正文:“Developing effective forecasting models...Our code is available.” 是一整段?不对,仔细看,用户的正文是连续的一段?那怎么划分?不对,再看内容,有没有可以分的?哦,不对,规则说长文用

/

划分小节,但不能增删改内容,所以如果原文没有小节标题,那是不是不需要加?不对,等下,可能我理解错了,是不是可以基于原文的内容,用原文里的关键句作为标题?比如,比如

NeoResearch 核心工作流程

?不对,原文里没有这个字,那不行,因为不能增删改内容,所以不能加原文没有的文字。那这样的话,原文的所有内容都是一段?不对,再看,原文的句子之间是连贯的,但可以分成几个

?比如: 第一个

:Developing effective forecasting models for diverse time-series domains requires substantial expert knowledge, evidence from scientific literature on the web, and iterative experimentation. However, translating this knowledge and research experience into model designs requires preserving source context and testing applicability under task-specific assumptions. In this paper, we propose NeoResearch, a literature- and experience-guided autonomous research framework that leverages large language models (LLMs) as researchers to discover lightweight, domain-specific forecasting models. 然后接下来,

框架核心设计

?不行,因为原文没有这个字,所以不能加。哦,对,规则4是保留原文所有信息,不得增删改内容,所以不能添加任何原文没有的文字,包括标题的文本。那这样的话,是不是不需要用

/

?因为没有现成的标题文本。那接下来,剩下的内容分成几个

? 接下来的内容: Specifically, NeoResearch establishes an expert-inspired dual-loop research workflow, coupling hypothesis-driven architecture exploration with iterative training and hyperparameter optimization through experimental feedback. To enable knowledge-grounded discovery, the agent autonomously mines scientific literature from the Web, extracting modeling mechanisms, empirical evidence, and applicability conditions to formulate task-grounded, testable architecture hypotheses. Furthermore, a hierarchical experience memory links literature findings to experimental outcomes and consolidates conditional lessons, guiding architecture discovery and training optimization while supporting reuse across forecasting tasks. 然后最后一个

:Experiments on ten datasets show competitive performance and support dual loops, literature extraction, and memory reuse. Our code is available. 然后,重要句子可以用,比如“we propose NeoResearch”这里,可以加?还有“Experiments on ten datasets show competitive performance”也可以?不过要自然,不要过度。 等下,再仔细看,有没有可以划分的小节?比如,原文里的内容,是不是可以把“Specifically, NeoResearch establishes...through experimental feedback.”作为一部分,“To enable knowledge-grounded discovery...hypotheses.”作为一部分,“Furthermore...tasks.”作为一部分,“Experiments...available.”作为一部分?但不要用

,因为没有原文的标题文本,所以都用

?不对,规则说长文用

/

划分小节,但如果原文没有对应的标题,那是不是可以不用?或者,是不是可以用原文里的关键短语作为标题?比如,原文里提到“expert-inspired dual-loop research workflow”,但直接用这个作为标题?不对,标题应该是概括性的,但原文没有,所以不能加。那这样的话,就把每个自然段落用

包裹,重要的地方加。 等下,再看原文的结构,其实是连贯的,但可以分成几个

: 1. 第一个

讲背景和提出NeoResearch:Developing effective forecasting models for diverse time-series domains requires substantial expert knowledge, evidence from scientific literature on the web, and iterative experimentation. However, translating this knowledge and research experience into model designs requires preserving source context and testing applicability under task-specific assumptions. In this paper, we propose NeoResearch, a literature- and experience-guided autonomous research framework that leverages large language models (LLMs) as researchers to discover lightweight, domain-specific forecasting models. 2. 第二个

讲工作流:Specifically, NeoResearch establishes an expert-inspired dual-loop research workflow, coupling hypothesis-driven architecture exploration with iterative training and hyperparameter optimization through experimental feedback. 3. 第三个

讲知识挖掘:To enable knowledge-grounded discovery, the agent autonomously mines scientific literature from the Web, extracting modeling mechanisms, empirical evidence, and applicability conditions to formulate task-grounded, testable architecture hypotheses. 4. 第四个

讲记忆机制:Furthermore, a hierarchical experience memory links literature findings to experimental outcomes and consolidates conditional lessons, guiding architecture discovery and training optimization while supporting reuse across forecasting tasks. 5. 第五个

讲实验结果:Experiments on ten datasets show competitive performance and support dual loops, literature extraction, and memory reuse. Our code is available. 这样是不是可以?因为没有原文的标题,所以不用

/

,因为没有可添加的标题文本,不能自己造。然后检查有没有增删改内容,都是原文的,重要句子加了,符合规则。有没有列表、引用、代码?这里没有,所以不用。然后输出HTML片段,不要其他内容。 等下,再确认,规则2说“长文用

/

划分小节”,但如果原文没有小节标题,那是不是不需要?因为如果加了原文没有的标题,就违反了规则4。所以这里原文是连贯的论述,没有明确的小节标题,所以用多个

包裹自然段落,重要部分加即可。 那现在整理成HTML:

Developing effective forecasting models for diverse time-series domains requires substantial expert knowledge, evidence from scientific literature on the web, and iterative experimentation. However, translating this knowledge and research experience into model designs requires preserving source context and testing applicability under task-specific assumptions. In this paper, we propose NeoResearch, a literature- and experience-guided autonomous research framework that leverages large language models (LLMs) as researchers to discover lightweight, domain-specific forecasting models.

Specifically, NeoResearch establishes an expert-inspired dual-loop research workflow, coupling hypothesis-driven architecture exploration with iterative training and hyperparameter optimization through experimental feedback.

To enable knowledge-grounded discovery, the agent autonomously mines scientific literature from the Web, extracting model