Saved in:
Bibliographic Details
Main Authors: Zhang, Collin, Morris, John X., Shmatikov, Vitaly
Format: Preprint
Published: 2024
Subjects:
Online Access:https://arxiv.org/abs/2405.15012
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866910640144121856
author Zhang, Collin
Morris, John X.
Shmatikov, Vitaly
author_facet Zhang, Collin
Morris, John X.
Shmatikov, Vitaly
contents We consider the problem of language model inversion: given outputs of a language model, we seek to extract the prompt that generated these outputs. We develop a new black-box method, output2prompt, that learns to extract prompts without access to the model's logits and without adversarial or jailbreaking queries. In contrast to previous work, output2prompt only needs outputs of normal user queries. To improve memory efficiency, output2prompt employs a new sparse encoding techique. We measure the efficacy of output2prompt on a variety of user and system prompts and demonstrate zero-shot transferability across different LLMs.
format Preprint
id arxiv_https___arxiv_org_abs_2405_15012
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Extracting Prompts by Inverting LLM Outputs
Zhang, Collin
Morris, John X.
Shmatikov, Vitaly
Computation and Language
Machine Learning
We consider the problem of language model inversion: given outputs of a language model, we seek to extract the prompt that generated these outputs. We develop a new black-box method, output2prompt, that learns to extract prompts without access to the model's logits and without adversarial or jailbreaking queries. In contrast to previous work, output2prompt only needs outputs of normal user queries. To improve memory efficiency, output2prompt employs a new sparse encoding techique. We measure the efficacy of output2prompt on a variety of user and system prompts and demonstrate zero-shot transferability across different LLMs.
title Extracting Prompts by Inverting LLM Outputs
topic Computation and Language
Machine Learning
url https://arxiv.org/abs/2405.15012