Matryoshka Pilot: Learning to Drive Black-Box LLMs with LLMs

Fuente: arXiv
Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Li, Changhao, Zhuang, Yuchen, Qiang, Rushi, Sun, Haotian, Dai, Hanjun, Zhang, Chao, Dai, Bo
Format: Preprint
Veröffentlicht: 2024
Schlagworte:
Online-Zugang:
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
_version_ 1866911262591418368
author Li, Changhao
Zhuang, Yuchen
Qiang, Rushi
Sun, Haotian
Dai, Hanjun
Zhang, Chao
Dai, Bo
author_facet Li, Changhao
Zhuang, Yuchen
Qiang, Rushi
Sun, Haotian
Dai, Hanjun
Zhang, Chao
Dai, Bo
contents Despite the impressive generative abilities of black-box large language models (LLMs), their inherent opacity hinders further advancements in capabilities such as reasoning, planning, and personalization. Existing works aim to enhance LLM capabilities via domain-specific adaptation, which require additional training on accessible model parameters, an infeasible option for black-box LLMs. To address this challenge, we introduce Matryoshka Pilot (M-Pilot), a lightweight white-box LLM controller that guides a large-scale black-box LLM generator by decomposing complex tasks into a series of intermediate outputs. Specifically, we consider the black-box LLM as an environment, with M-Pilot serving as a policy to provide intermediate guidance through prompts for driving the black-box LLM. M-Pilot is trained to pivot the outputs of the black-box LLM aligning with preferences during iterative interaction, which enables controllable multi-turn generation and self-improvement in optimizing intermediate guidance. Empirical evaluations on diverse tasks demonstrate that our method effectively enhances the capabilities of black-box LLMs in complex, long-horizon tasks. Our code is publicly available at: https://github.com/lichangh20/Matryoshka.
format Preprint
id arxiv_https___arxiv_org_abs_2410_20749
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Matryoshka Pilot: Learning to Drive Black-Box LLMs with LLMs
Li, Changhao
Zhuang, Yuchen
Qiang, Rushi
Sun, Haotian
Dai, Hanjun
Zhang, Chao
Dai, Bo
Machine Learning
Artificial Intelligence
Computation and Language
Despite the impressive generative abilities of black-box large language models (LLMs), their inherent opacity hinders further advancements in capabilities such as reasoning, planning, and personalization. Existing works aim to enhance LLM capabilities via domain-specific adaptation, which require additional training on accessible model parameters, an infeasible option for black-box LLMs. To address this challenge, we introduce Matryoshka Pilot (M-Pilot), a lightweight white-box LLM controller that guides a large-scale black-box LLM generator by decomposing complex tasks into a series of intermediate outputs. Specifically, we consider the black-box LLM as an environment, with M-Pilot serving as a policy to provide intermediate guidance through prompts for driving the black-box LLM. M-Pilot is trained to pivot the outputs of the black-box LLM aligning with preferences during iterative interaction, which enables controllable multi-turn generation and self-improvement in optimizing intermediate guidance. Empirical evaluations on diverse tasks demonstrate that our method effectively enhances the capabilities of black-box LLMs in complex, long-horizon tasks. Our code is publicly available at: https://github.com/lichangh20/Matryoshka.
title Matryoshka Pilot: Learning to Drive Black-Box LLMs with LLMs
topic Machine Learning
Artificial Intelligence
Computation and Language
url https://arxiv.org/abs/2410.20749