Guardado en:
Detalles Bibliográficos
Autores principales: Li, Zongyuan, Li, Pengfei, Qi, Runnan, Ni, Yanan, Jiang, Lumin, Wu, Hui, Zhang, Xuebo, Huang, Kuihua, Guo, Xian
Formato: Preprint
Publicado: 2025
Materias:
Acceso en línea:https://arxiv.org/abs/2505.01073
Etiquetas: Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
_version_ 1866908346201669632
author Li, Zongyuan
Li, Pengfei
Qi, Runnan
Ni, Yanan
Jiang, Lumin
Wu, Hui
Zhang, Xuebo
Huang, Kuihua
Guo, Xian
author_facet Li, Zongyuan
Li, Pengfei
Qi, Runnan
Ni, Yanan
Jiang, Lumin
Wu, Hui
Zhang, Xuebo
Huang, Kuihua
Guo, Xian
contents The lack of domain-specific data in the pre-training of Large Language Models (LLMs) severely limits LLM-based decision systems in specialized applications, while post-training a model in the scenarios requires significant computational resources. In this paper, we present Retrial-Augmented Learning (RAL), a reward-free self-supervised learning framework for LLMs that operates without model training. By developing Retrieval-Augmented Generation (RAG) into a module for organizing intermediate data, we realized a three-stage autonomous knowledge generation of proposing a hypothesis, validating the hypothesis, and generating the knowledge. The method is evaluated in the LLM-PySC2 environment, a representative decision-making platform that combines sufficient complexity with domain-specific knowledge requirements. Experiments demonstrate that the proposed method effectively reduces hallucination by generating and utilizing validated knowledge, and increases decision-making performance at an extremely low cost. Meanwhile, the approach exhibits potential in out-of-distribution(OOD) tasks, robustness, and transferability, making it a cost-friendly but effective solution for decision-making problems and autonomous knowledge generation.
format Preprint
id arxiv_https___arxiv_org_abs_2505_01073
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Retrieval Augmented Learning: A Retrial-based Large Language Model Self-Supervised Learning and Autonomous Knowledge Generation
Li, Zongyuan
Li, Pengfei
Qi, Runnan
Ni, Yanan
Jiang, Lumin
Wu, Hui
Zhang, Xuebo
Huang, Kuihua
Guo, Xian
Artificial Intelligence
The lack of domain-specific data in the pre-training of Large Language Models (LLMs) severely limits LLM-based decision systems in specialized applications, while post-training a model in the scenarios requires significant computational resources. In this paper, we present Retrial-Augmented Learning (RAL), a reward-free self-supervised learning framework for LLMs that operates without model training. By developing Retrieval-Augmented Generation (RAG) into a module for organizing intermediate data, we realized a three-stage autonomous knowledge generation of proposing a hypothesis, validating the hypothesis, and generating the knowledge. The method is evaluated in the LLM-PySC2 environment, a representative decision-making platform that combines sufficient complexity with domain-specific knowledge requirements. Experiments demonstrate that the proposed method effectively reduces hallucination by generating and utilizing validated knowledge, and increases decision-making performance at an extremely low cost. Meanwhile, the approach exhibits potential in out-of-distribution(OOD) tasks, robustness, and transferability, making it a cost-friendly but effective solution for decision-making problems and autonomous knowledge generation.
title Retrieval Augmented Learning: A Retrial-based Large Language Model Self-Supervised Learning and Autonomous Knowledge Generation
topic Artificial Intelligence
url https://arxiv.org/abs/2505.01073