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 2402.01620cs.CL
收藏

MAGDi:多智能体交互图的结构化蒸馏提升小型语言模型的推理能力

MAGDi: Structured Distillation of Multi-Agent Interaction Graphs Improves Reasoning in Smaller Language Models

UNC Chapel Hill(北卡罗来纳大学教堂山分校)

Justin Chih-Yao Chen, Swarnadeep Saha, Elias Stengel-Eskin, Mohit Bansal

AI总结:

本文提出 MAGDi 方法,将多智能体交互建模为图并对小型语言模型进行结构化蒸馏,在多个推理基准上超越单教师和多教师蒸馏方法,同时显著提升推理效率与泛化能力。

摘要:

大型语言模型(LLM)智能体之间的多智能体交互已在多种推理任务上取得显著改进。然而,这些方法需要在多轮中由多个模型生成长序列,成本高昂。此外,这些多智能体方法无法提供一个最终的单一模型用于高效推理。为解决这一问题,我们提出 MAGDi,一种将多个 LLM 之间的推理交互结构化蒸馏到较小语言模型(LM)中的新方法。MAGDi 通过将多智能体交互表示为图来教导较小的模型,为基础学生模型增加一个图编码器,并使用三个目标函数蒸馏知识:下一词元预测、正确与错误推理之间的对比损失,以及用于建模交互结构的基于图的目标。在七个广泛使用的常识和数学推理基准上的实验表明,MAGDi 提升了较小模型的推理能力,优于多种从单个教师和多个教师蒸馏的方法。此外,MAGDi 还展示出比其教师高一个数量级的效率。我们进行了广泛分析,表明 MAGDi(1)增强了对域外任务的泛化能力,(2)随着基础学生模型的规模和强度正向扩展,(3)在应用自洽性(self-consistency)——一种依赖模型多样性的推理技术——时(通过我们的多教师训练)获得更大改进。

英文摘要:

Multi-agent interactions between Large Language Model (LLM) agents have shown major improvements on diverse reasoning tasks. However, these involve long generations from multiple models across several rounds, making them expensive. Moreover, these multi-agent approaches fail to provide a final, single model for efficient inference. To address this, we introduce MAGDi, a new method for structured distillation of the reasoning interactions between multiple LLMs into smaller LMs. MAGDi teaches smaller models by representing multi-agent interactions as graphs, augmenting a base student model with a graph encoder, and distilling knowledge using three objective functions: next-token prediction, a contrastive loss between correct and incorrect reasoning, and a graph-based objective to model the interaction structure. Experiments on seven widely used commonsense and math reasoning benchmarks show that MAGDi improves the reasoning capabilities of smaller models, outperforming several methods that distill from a single teacher and multiple teachers. Moreover, MAGDi also demonstrates an order of magnitude higher efficiency over its teachers. We conduct extensive analyses to show that MAGDi (1) enhances the generalizability to out-of-domain tasks, (2) scales positively with the size and strength of the base student model, and (3) obtains larger improvements (via our multi-teacher training) when applying self-consistency -- an inference technique that relies on model diversity.

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。

↑