@inproceedings{bounhar-etal-2026-yapo,
title = "{Y}a{PO}: Learnable Sparse Activation Steering Vectors for Domain Adaptation",
author = "Bounhar, Abdelaziz and
Elbadry, Rania and
Abdine, Hadi and
Nakov, Preslav and
Vazirgiannis, Michalis and
Shang, Guokan",
editor = "Zhang, Yue and
Xu, Ruifeng and
Purwarianti, Ayu and
Rong, Tong and
Suzuki, Jun",
booktitle = "Findings of the {A}ssociation for {C}omputational {L}inguistics: {IJCNLP}-{AACL} 2026",
month = nov,
year = "2026",
address = "Hengqin, China",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2026.findings-aacl.240/",
pages = "4719--4740",
ISBN = "979-8-89176-444-6",
abstract = "Activation steering is a lightweight alternative to fine-tuning for aligning and personalizing large language models (LLMs). Prior work such as Bi-directional Preference Optimization (BiPO) learns dense steering vectors from preference data. However, dense interventions can entangle latent factors due to neuron multi-semanticity, hurting stability and fine-grained control, thus limiting their effectiveness especially in cultural alignment, where closely related values and behaviors (e.g., across Middle Eastern cultures) must be distinguished. We propose Yet another Policy Optimization (YaPO), a reference-free method that learns preference-optimized steering vectors in the sparse latent space of a Sparse Autoencoder (SAE). While the learned vector is not explicitly constrained to be sparse, it acts over sparse SAE features, providing a more disentangled and interpretable intervention basis. By optimizing sparse codes, YaPO yields more disentangled and interpretable steering directions. Across cultural alignment and several alignment-related behaviors (e.g., hallucination reduction and jailbreak resistance), YaPO converges faster and is more stable than dense steering baselines while improving performance. Importantly, YaPO preserves general knowledge, with no measurable degradation on MMLU. Overall, our results show that YaPO provides a general recipe for efficient, stable, and fine-grained alignment of LLMs, with broad applications to controllability and domain adaptation."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="bounhar-etal-2026-yapo">
<titleInfo>
<title>YaPO: Learnable Sparse Activation Steering Vectors for Domain Adaptation</title>
</titleInfo>
<name type="personal">
<namePart type="given">Abdelaziz</namePart>
<namePart type="family">Bounhar</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Rania</namePart>
<namePart type="family">Elbadry</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Hadi</namePart>
<namePart type="family">Abdine</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Preslav</namePart>
<namePart type="family">Nakov</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Michalis</namePart>
<namePart type="family">Vazirgiannis</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Guokan</namePart>
<namePart type="family">Shang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<originInfo>
<dateIssued>2026-11</dateIssued>
</originInfo>
<typeOfResource>text</typeOfResource>
<relatedItem type="host">
<titleInfo>
<title>Findings of the Association for Computational Linguistics: IJCNLP-AACL 2026</title>
</titleInfo>
<name type="personal">
<namePart type="given">Yue</namePart>
<namePart type="family">Zhang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Ruifeng</namePart>
<namePart type="family">Xu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Ayu</namePart>
<namePart type="family">Purwarianti</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Tong</namePart>
<namePart type="family">Rong</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Jun</namePart>
<namePart type="family">Suzuki</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<originInfo>
<publisher>Association for Computational Linguistics</publisher>
<place>
<placeTerm type="text">Hengqin, China</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
<identifier type="isbn">979-8-89176-444-6</identifier>
</relatedItem>
<abstract>Activation steering is a lightweight alternative to fine-tuning for aligning and personalizing large language models (LLMs). Prior work such as Bi-directional Preference Optimization (BiPO) learns dense steering vectors from preference data. However, dense interventions can entangle latent factors due to neuron multi-semanticity, hurting stability and fine-grained control, thus limiting their effectiveness especially in cultural alignment, where closely related values and behaviors (e.g., across Middle Eastern cultures) must be distinguished. We propose Yet another Policy Optimization (YaPO), a reference-free method that learns preference-optimized steering vectors in the sparse latent space of a Sparse Autoencoder (SAE). While the learned vector is not explicitly constrained to be sparse, it acts over sparse SAE features, providing a more disentangled and interpretable intervention basis. By optimizing sparse codes, YaPO yields more disentangled and interpretable steering directions. Across cultural alignment and several alignment-related behaviors (e.g., hallucination reduction and jailbreak resistance), YaPO converges faster and is more stable than dense steering baselines while improving performance. Importantly, YaPO preserves general knowledge, with no measurable degradation on MMLU. Overall, our results show that YaPO provides a general recipe for efficient, stable, and fine-grained alignment of LLMs, with broad applications to controllability and domain adaptation.</abstract>
<identifier type="citekey">bounhar-etal-2026-yapo</identifier>
<location>
<url>https://aclanthology.org/2026.findings-aacl.240/</url>
</location>
<part>
<date>2026-11</date>
<extent unit="page">
<start>4719</start>
<end>4740</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T YaPO: Learnable Sparse Activation Steering Vectors for Domain Adaptation
%A Bounhar, Abdelaziz
%A Elbadry, Rania
%A Abdine, Hadi
%A Nakov, Preslav
%A Vazirgiannis, Michalis
%A Shang, Guokan
%Y Zhang, Yue
%Y Xu, Ruifeng
%Y Purwarianti, Ayu
%Y Rong, Tong
%Y Suzuki, Jun
%S Findings of the Association for Computational Linguistics: IJCNLP-AACL 2026
%D 2026
%8 November
%I Association for Computational Linguistics
%C Hengqin, China
%@ 979-8-89176-444-6
%F bounhar-etal-2026-yapo
%X Activation steering is a lightweight alternative to fine-tuning for aligning and personalizing large language models (LLMs). Prior work such as Bi-directional Preference Optimization (BiPO) learns dense steering vectors from preference data. However, dense interventions can entangle latent factors due to neuron multi-semanticity, hurting stability and fine-grained control, thus limiting their effectiveness especially in cultural alignment, where closely related values and behaviors (e.g., across Middle Eastern cultures) must be distinguished. We propose Yet another Policy Optimization (YaPO), a reference-free method that learns preference-optimized steering vectors in the sparse latent space of a Sparse Autoencoder (SAE). While the learned vector is not explicitly constrained to be sparse, it acts over sparse SAE features, providing a more disentangled and interpretable intervention basis. By optimizing sparse codes, YaPO yields more disentangled and interpretable steering directions. Across cultural alignment and several alignment-related behaviors (e.g., hallucination reduction and jailbreak resistance), YaPO converges faster and is more stable than dense steering baselines while improving performance. Importantly, YaPO preserves general knowledge, with no measurable degradation on MMLU. Overall, our results show that YaPO provides a general recipe for efficient, stable, and fine-grained alignment of LLMs, with broad applications to controllability and domain adaptation.
%U https://aclanthology.org/2026.findings-aacl.240/
%P 4719-4740
Markdown (Informal)
[YaPO: Learnable Sparse Activation Steering Vectors for Domain Adaptation](https://aclanthology.org/2026.findings-aacl.240/) (Bounhar et al., Findings 2026)
ACL