@inproceedings{shin-etal-2025-generative,
title = "Generative Prompt Internalization",
author = "Shin, Haebin and
Ji, Lei and
Gong, Yeyun and
Kim, Sungdong and
Choi, Eunbi and
Seo, Minjoon",
editor = "Chiruzzo, Luis and
Ritter, Alan and
Wang, Lu",
booktitle = "Proceedings of the 2025 Conference of the Nations of the Americas Chapter of the Association for Computational Linguistics: Human Language Technologies (Volume 1: Long Papers)",
month = apr,
year = "2025",
address = "Albuquerque, New Mexico",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2025.naacl-long.376/",
doi = "10.18653/v1/2025.naacl-long.376",
pages = "7338--7363",
ISBN = "979-8-89176-189-6",
abstract = "Prompts used in recent large language model based applications are often fixed and lengthy, leading to significant computational overhead. To address this challenge, we propose Generative Prompt Internalization (GenPI), a lightweight method that employs a joint training approach. GenPI not only replicates the behavior of models with prompt inputs but also generates the content of the prompt along with reasons for why the model{'}s behavior should change accordingly. We demonstrate that our approach effectively internalizes complex prompts across various agent-based application scenarios. For effective training without interactions with the dedicated environments, we introduce a data synthesis technique that autonomously collects conversational datasets by swapping the roles of the agent and environment. This method is especially useful in scenarios where only a predefined prompt is available without a corresponding training dataset. By internalizing complex prompts, Generative Prompt Internalization enables high performance and efficient inference without the need for explicit prompts."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="shin-etal-2025-generative">
<titleInfo>
<title>Generative Prompt Internalization</title>
</titleInfo>
<name type="personal">
<namePart type="given">Haebin</namePart>
<namePart type="family">Shin</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Lei</namePart>
<namePart type="family">Ji</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yeyun</namePart>
<namePart type="family">Gong</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Sungdong</namePart>
<namePart type="family">Kim</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Eunbi</namePart>
<namePart type="family">Choi</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Minjoon</namePart>
<namePart type="family">Seo</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<originInfo>
<dateIssued>2025-04</dateIssued>
</originInfo>
<typeOfResource>text</typeOfResource>
<relatedItem type="host">
<titleInfo>
<title>Proceedings of the 2025 Conference of the Nations of the Americas Chapter of the Association for Computational Linguistics: Human Language Technologies (Volume 1: Long Papers)</title>
</titleInfo>
<name type="personal">
<namePart type="given">Luis</namePart>
<namePart type="family">Chiruzzo</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Alan</namePart>
<namePart type="family">Ritter</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Lu</namePart>
<namePart type="family">Wang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<originInfo>
<publisher>Association for Computational Linguistics</publisher>
<place>
<placeTerm type="text">Albuquerque, New Mexico</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
<identifier type="isbn">979-8-89176-189-6</identifier>
</relatedItem>
<abstract>Prompts used in recent large language model based applications are often fixed and lengthy, leading to significant computational overhead. To address this challenge, we propose Generative Prompt Internalization (GenPI), a lightweight method that employs a joint training approach. GenPI not only replicates the behavior of models with prompt inputs but also generates the content of the prompt along with reasons for why the model’s behavior should change accordingly. We demonstrate that our approach effectively internalizes complex prompts across various agent-based application scenarios. For effective training without interactions with the dedicated environments, we introduce a data synthesis technique that autonomously collects conversational datasets by swapping the roles of the agent and environment. This method is especially useful in scenarios where only a predefined prompt is available without a corresponding training dataset. By internalizing complex prompts, Generative Prompt Internalization enables high performance and efficient inference without the need for explicit prompts.</abstract>
<identifier type="citekey">shin-etal-2025-generative</identifier>
<identifier type="doi">10.18653/v1/2025.naacl-long.376</identifier>
<location>
<url>https://aclanthology.org/2025.naacl-long.376/</url>
</location>
<part>
<date>2025-04</date>
<extent unit="page">
<start>7338</start>
<end>7363</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Generative Prompt Internalization
%A Shin, Haebin
%A Ji, Lei
%A Gong, Yeyun
%A Kim, Sungdong
%A Choi, Eunbi
%A Seo, Minjoon
%Y Chiruzzo, Luis
%Y Ritter, Alan
%Y Wang, Lu
%S Proceedings of the 2025 Conference of the Nations of the Americas Chapter of the Association for Computational Linguistics: Human Language Technologies (Volume 1: Long Papers)
%D 2025
%8 April
%I Association for Computational Linguistics
%C Albuquerque, New Mexico
%@ 979-8-89176-189-6
%F shin-etal-2025-generative
%X Prompts used in recent large language model based applications are often fixed and lengthy, leading to significant computational overhead. To address this challenge, we propose Generative Prompt Internalization (GenPI), a lightweight method that employs a joint training approach. GenPI not only replicates the behavior of models with prompt inputs but also generates the content of the prompt along with reasons for why the model’s behavior should change accordingly. We demonstrate that our approach effectively internalizes complex prompts across various agent-based application scenarios. For effective training without interactions with the dedicated environments, we introduce a data synthesis technique that autonomously collects conversational datasets by swapping the roles of the agent and environment. This method is especially useful in scenarios where only a predefined prompt is available without a corresponding training dataset. By internalizing complex prompts, Generative Prompt Internalization enables high performance and efficient inference without the need for explicit prompts.
%R 10.18653/v1/2025.naacl-long.376
%U https://aclanthology.org/2025.naacl-long.376/
%U https://doi.org/10.18653/v1/2025.naacl-long.376
%P 7338-7363
Markdown (Informal)
[Generative Prompt Internalization](https://aclanthology.org/2025.naacl-long.376/) (Shin et al., NAACL 2025)
ACL
- Haebin Shin, Lei Ji, Yeyun Gong, Sungdong Kim, Eunbi Choi, and Minjoon Seo. 2025. Generative Prompt Internalization. In Proceedings of the 2025 Conference of the Nations of the Americas Chapter of the Association for Computational Linguistics: Human Language Technologies (Volume 1: Long Papers), pages 7338–7363, Albuquerque, New Mexico. Association for Computational Linguistics.