Guardado en:
Detalles Bibliográficos
Autores principales: Ji, Miaomiao, Wu, Yanqiu, Wu, Zhibin, Wang, Shoujin, Yang, Jian, Dras, Mark, Naseem, Usman
Formato: Preprint
Publicado: 2025
Materias:
Acceso en línea:https://arxiv.org/abs/2505.02666
Etiquetas: Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
_version_ 1866915473858232320
author Ji, Miaomiao
Wu, Yanqiu
Wu, Zhibin
Wang, Shoujin
Yang, Jian
Dras, Mark
Naseem, Usman
author_facet Ji, Miaomiao
Wu, Yanqiu
Wu, Zhibin
Wang, Shoujin
Yang, Jian
Dras, Mark
Naseem, Usman
contents Reward design plays a pivotal role in aligning large language models (LLMs) with human values, serving as the bridge between feedback signals and model optimization. This survey provides a structured organization of reward modeling and addresses three key aspects: mathematical formulation, construction practices, and interaction with optimization paradigms. Building on this, it develops a macro-level taxonomy that characterizes reward mechanisms along complementary dimensions, thereby offering both conceptual clarity and practical guidance for alignment research. The progression of LLM alignment can be understood as a continuous refinement of reward design strategies, with recent developments highlighting paradigm shifts from reinforcement learning (RL)-based to RL-free optimization and from single-task to multi-objective and complex settings.
format Preprint
id arxiv_https___arxiv_org_abs_2505_02666
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle A Survey on Progress in LLM Alignment from the Perspective of Reward Design
Ji, Miaomiao
Wu, Yanqiu
Wu, Zhibin
Wang, Shoujin
Yang, Jian
Dras, Mark
Naseem, Usman
Computation and Language
Reward design plays a pivotal role in aligning large language models (LLMs) with human values, serving as the bridge between feedback signals and model optimization. This survey provides a structured organization of reward modeling and addresses three key aspects: mathematical formulation, construction practices, and interaction with optimization paradigms. Building on this, it develops a macro-level taxonomy that characterizes reward mechanisms along complementary dimensions, thereby offering both conceptual clarity and practical guidance for alignment research. The progression of LLM alignment can be understood as a continuous refinement of reward design strategies, with recent developments highlighting paradigm shifts from reinforcement learning (RL)-based to RL-free optimization and from single-task to multi-objective and complex settings.
title A Survey on Progress in LLM Alignment from the Perspective of Reward Design
topic Computation and Language
url https://arxiv.org/abs/2505.02666