@inproceedings{dong-etal-2025-contrans,
title = "{CONTRANS}: Weak-to-Strong Alignment Engineering via Concept Transplantation",
author = "Dong, Weilong and
Wu, Xinwei and
Jin, Renren and
Xu, Shaoyang and
Xiong, Deyi",
editor = "Rambow, Owen and
Wanner, Leo and
Apidianaki, Marianna and
Al-Khalifa, Hend and
Eugenio, Barbara Di and
Schockaert, Steven",
booktitle = "Proceedings of the 31st International Conference on Computational Linguistics",
month = jan,
year = "2025",
address = "Abu Dhabi, UAE",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2025.coling-main.279/",
pages = "4130--4148",
abstract = "Ensuring large language models (LLM) behave consistently with human goals, values, and intentions is crucial for their safety but yet computationally expensive. To reduce the computational cost of alignment training of LLMs, especially for those with a huge number of parameters, and to reutilize learned value alignment, we propose ConTrans, a novel framework that enables weak-to-strong alignment transfer via concept transplantation. From the perspective of representation engineering, ConTrans refines concept vectors in value alignment from a source LLM (usually a weak yet aligned LLM). The refined concept vectors are then reformulated to adapt to the target LLM (usually a strong yet unaligned base LLM) via affine transformation. In the third step, ConTrans transplants the reformulated concept vectors into the residual stream of the target LLM. Experiments demonstrate the successful transplantation of a wide range of aligned concepts from 7B models to 13B and 70B models across multiple LLMs and LLM families. Remarkably, ConTrans even surpasses instruction-tuned models in terms of truthfulness. Experiment results validate the effectiveness of both inter-LLM-family and intra-LLM-family concept transplantation. Our work successfully demonstrates an alternative way to achieve weak-to-strong alignment generalization and control."
}
<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="dong-etal-2025-contrans">
<titleInfo>
<title>CONTRANS: Weak-to-Strong Alignment Engineering via Concept Transplantation</title>
</titleInfo>
<name type="personal">
<namePart type="given">Weilong</namePart>
<namePart type="family">Dong</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Xinwei</namePart>
<namePart type="family">Wu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Renren</namePart>
<namePart type="family">Jin</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Shaoyang</namePart>
<namePart type="family">Xu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Deyi</namePart>
<namePart type="family">Xiong</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<originInfo>
<dateIssued>2025-01</dateIssued>
</originInfo>
<typeOfResource>text</typeOfResource>
<relatedItem type="host">
<titleInfo>
<title>Proceedings of the 31st International Conference on Computational Linguistics</title>
</titleInfo>
<name type="personal">
<namePart type="given">Owen</namePart>
<namePart type="family">Rambow</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Leo</namePart>
<namePart type="family">Wanner</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Marianna</namePart>
<namePart type="family">Apidianaki</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Hend</namePart>
<namePart type="family">Al-Khalifa</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Barbara</namePart>
<namePart type="given">Di</namePart>
<namePart type="family">Eugenio</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Steven</namePart>
<namePart type="family">Schockaert</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<originInfo>
<publisher>Association for Computational Linguistics</publisher>
<place>
<placeTerm type="text">Abu Dhabi, UAE</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
</relatedItem>
<abstract>Ensuring large language models (LLM) behave consistently with human goals, values, and intentions is crucial for their safety but yet computationally expensive. To reduce the computational cost of alignment training of LLMs, especially for those with a huge number of parameters, and to reutilize learned value alignment, we propose ConTrans, a novel framework that enables weak-to-strong alignment transfer via concept transplantation. From the perspective of representation engineering, ConTrans refines concept vectors in value alignment from a source LLM (usually a weak yet aligned LLM). The refined concept vectors are then reformulated to adapt to the target LLM (usually a strong yet unaligned base LLM) via affine transformation. In the third step, ConTrans transplants the reformulated concept vectors into the residual stream of the target LLM. Experiments demonstrate the successful transplantation of a wide range of aligned concepts from 7B models to 13B and 70B models across multiple LLMs and LLM families. Remarkably, ConTrans even surpasses instruction-tuned models in terms of truthfulness. Experiment results validate the effectiveness of both inter-LLM-family and intra-LLM-family concept transplantation. Our work successfully demonstrates an alternative way to achieve weak-to-strong alignment generalization and control.</abstract>
<identifier type="citekey">dong-etal-2025-contrans</identifier>
<location>
<url>https://aclanthology.org/2025.coling-main.279/</url>
</location>
<part>
<date>2025-01</date>
<extent unit="page">
<start>4130</start>
<end>4148</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T CONTRANS: Weak-to-Strong Alignment Engineering via Concept Transplantation
%A Dong, Weilong
%A Wu, Xinwei
%A Jin, Renren
%A Xu, Shaoyang
%A Xiong, Deyi
%Y Rambow, Owen
%Y Wanner, Leo
%Y Apidianaki, Marianna
%Y Al-Khalifa, Hend
%Y Eugenio, Barbara Di
%Y Schockaert, Steven
%S Proceedings of the 31st International Conference on Computational Linguistics
%D 2025
%8 January
%I Association for Computational Linguistics
%C Abu Dhabi, UAE
%F dong-etal-2025-contrans
%X Ensuring large language models (LLM) behave consistently with human goals, values, and intentions is crucial for their safety but yet computationally expensive. To reduce the computational cost of alignment training of LLMs, especially for those with a huge number of parameters, and to reutilize learned value alignment, we propose ConTrans, a novel framework that enables weak-to-strong alignment transfer via concept transplantation. From the perspective of representation engineering, ConTrans refines concept vectors in value alignment from a source LLM (usually a weak yet aligned LLM). The refined concept vectors are then reformulated to adapt to the target LLM (usually a strong yet unaligned base LLM) via affine transformation. In the third step, ConTrans transplants the reformulated concept vectors into the residual stream of the target LLM. Experiments demonstrate the successful transplantation of a wide range of aligned concepts from 7B models to 13B and 70B models across multiple LLMs and LLM families. Remarkably, ConTrans even surpasses instruction-tuned models in terms of truthfulness. Experiment results validate the effectiveness of both inter-LLM-family and intra-LLM-family concept transplantation. Our work successfully demonstrates an alternative way to achieve weak-to-strong alignment generalization and control.
%U https://aclanthology.org/2025.coling-main.279/
%P 4130-4148
Markdown (Informal)
[CONTRANS: Weak-to-Strong Alignment Engineering via Concept Transplantation](https://aclanthology.org/2025.coling-main.279/) (Dong et al., COLING 2025)
ACL