arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~
arXiv 2603.11583cs.CLcs.AI
收藏

UtilityMax Prompting:多目标大语言模型任务的形式化框架

UtilityMax Prompting: A Formal Framework for Multi-Objective Large Language Model Tasks

Independent Researcher(独立研究者)

Ofir Marom

AI总结:

提出UtilityMax Prompting框架,用影响图和期望效用最大化将多目标LLM任务形式化,在MovieLens 1M数据集上相比自然语言基线提升了精度和NDCG。

摘要:

大语言模型(LLM)任务的成功在很大程度上取决于其提示词。大多数用例使用自然语言指定提示词,当必须同时满足多个目标时,自然语言本质上是模糊的。在本文中,我们引入了UtilityMax Prompting,一个使用形式化数学语言指定任务的框架。我们将任务重构为一个影响图,其中LLM的答案是唯一的决策变量。在图中条件概率分布上定义效用函数,并指示LLM找到最大化期望效用的答案。这迫使LLM明确推理目标的每个组成部分,将其输出导向精确的优化目标,而非主观的自然语言解释。我们在MovieLens 1M数据集上,使用三个前沿模型(Claude Sonnet 4.6、GPT-5.4和Gemini 2.5 Pro)验证了我们的方法,在多目标电影推荐任务中,与自然语言基线相比,在精度和归一化折损累计增益(NDCG)上表现出一致的改进。

英文摘要:

The success of a Large Language Model (LLM) task depends heavily on its prompt. Most use-cases specify prompts using natural language, which is inherently ambiguous when multiple objectives must be simultaneously satisfied. In this paper we introduce UtilityMax Prompting, a framework that specifies tasks using formal mathematical language. We reconstruct the task as an influence diagram in which the LLM's answer is the sole decision variable. A utility function is defined over the conditional probability distributions within the diagram, and the LLM is instructed to find the answer that maximises expected utility. This constrains the LLM to reason explicitly about each component of the objective, directing its output toward a precise optimization target rather than a subjective natural language interpretation. We validate our approach on the MovieLens 1M dataset across three frontier models (Claude Sonnet 4.6, GPT-5.4, and Gemini 2.5 Pro), demonstrating consistent improvements in precision and Normalized Discounted Cumulative Gain (NDCG) over natural language baselines in a multi-objective movie recommendation task.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑