@inproceedings{mashayekhi-etal-2026-stress,
title = "Stress-Testing Random-Vector Controls for Activation Steering",
author = "Mashayekhi, Behnood and
Hosseini, Parsa and
Amiramjadi, Arian",
editor = "Arad, Dana and
Belinkov, Yonatan and
Chen, Hanjie and
Feldhus, Nils and
Mueller, Aaron and
Mohebbi, Hosein and
Sarti, Gabriele and
Tutek, Martin",
booktitle = "Proceedings of the 9th {B}lackbox{NLP} Workshop: Analyzing and Interpreting Neural Networks for {NLP}",
month = oct,
year = "2026",
address = "Budapest, Hungary",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2026.blackboxnlp-1.73/",
pages = "1264--1272",
ISBN = "979-8-89176-476-7",
abstract = "Random-vector controls provide a reference for interpreting activation-steering effects, but a single realization hides variation in that reference. We examine this variation and two separate implementation checks. In a released activation-addition (ActAdd) configuration on GPT-2-XL, we reproduce the steering arm over 57 prompts with a maximum absolute error of 1.74 {\texttimes} 10{\ensuremath{^-}}{\ensuremath{^6}}. At coefficient 10, 9 of 20 fresh random-control realizations have greater mean normalized Kullback{--}Leibler (KL) divergence than the steering arm, and 19 have lower means than the stored random realization. In a separate four-model study over 96 prompts, changing batch size or attention implementation changes the greedy next token for at most 5 and 6 prompts, respectively; an identical rerun changes none. Finally, an erroneous left-padding index in our own analysis pipeline, shared by the intervention and readout, overstates the maximum greedy-token change rate in 10 of 12 model-dose settings, with a median relative distortion of 46.8{\%}. These findings support reporting realization-level variation, measuring configuration sensitivity, and validating intervention and readout indices. The results concern distributional change in the evaluated configurations, not semantic steering success."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="mashayekhi-etal-2026-stress">
<titleInfo>
<title>Stress-Testing Random-Vector Controls for Activation Steering</title>
</titleInfo>
<name type="personal">
<namePart type="given">Behnood</namePart>
<namePart type="family">Mashayekhi</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Parsa</namePart>
<namePart type="family">Hosseini</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Arian</namePart>
<namePart type="family">Amiramjadi</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<originInfo>
<dateIssued>2026-10</dateIssued>
</originInfo>
<typeOfResource>text</typeOfResource>
<relatedItem type="host">
<titleInfo>
<title>Proceedings of the 9th BlackboxNLP Workshop: Analyzing and Interpreting Neural Networks for NLP</title>
</titleInfo>
<name type="personal">
<namePart type="given">Dana</namePart>
<namePart type="family">Arad</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yonatan</namePart>
<namePart type="family">Belinkov</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Hanjie</namePart>
<namePart type="family">Chen</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Nils</namePart>
<namePart type="family">Feldhus</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Aaron</namePart>
<namePart type="family">Mueller</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Hosein</namePart>
<namePart type="family">Mohebbi</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Gabriele</namePart>
<namePart type="family">Sarti</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Martin</namePart>
<namePart type="family">Tutek</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<originInfo>
<publisher>Association for Computational Linguistics</publisher>
<place>
<placeTerm type="text">Budapest, Hungary</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
<identifier type="isbn">979-8-89176-476-7</identifier>
</relatedItem>
<abstract>Random-vector controls provide a reference for interpreting activation-steering effects, but a single realization hides variation in that reference. We examine this variation and two separate implementation checks. In a released activation-addition (ActAdd) configuration on GPT-2-XL, we reproduce the steering arm over 57 prompts with a maximum absolute error of 1.74 × 10\ensuremath⁻\ensuremath⁶. At coefficient 10, 9 of 20 fresh random-control realizations have greater mean normalized Kullback–Leibler (KL) divergence than the steering arm, and 19 have lower means than the stored random realization. In a separate four-model study over 96 prompts, changing batch size or attention implementation changes the greedy next token for at most 5 and 6 prompts, respectively; an identical rerun changes none. Finally, an erroneous left-padding index in our own analysis pipeline, shared by the intervention and readout, overstates the maximum greedy-token change rate in 10 of 12 model-dose settings, with a median relative distortion of 46.8%. These findings support reporting realization-level variation, measuring configuration sensitivity, and validating intervention and readout indices. The results concern distributional change in the evaluated configurations, not semantic steering success.</abstract>
<identifier type="citekey">mashayekhi-etal-2026-stress</identifier>
<location>
<url>https://aclanthology.org/2026.blackboxnlp-1.73/</url>
</location>
<part>
<date>2026-10</date>
<extent unit="page">
<start>1264</start>
<end>1272</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Stress-Testing Random-Vector Controls for Activation Steering
%A Mashayekhi, Behnood
%A Hosseini, Parsa
%A Amiramjadi, Arian
%Y Arad, Dana
%Y Belinkov, Yonatan
%Y Chen, Hanjie
%Y Feldhus, Nils
%Y Mueller, Aaron
%Y Mohebbi, Hosein
%Y Sarti, Gabriele
%Y Tutek, Martin
%S Proceedings of the 9th BlackboxNLP Workshop: Analyzing and Interpreting Neural Networks for NLP
%D 2026
%8 October
%I Association for Computational Linguistics
%C Budapest, Hungary
%@ 979-8-89176-476-7
%F mashayekhi-etal-2026-stress
%X Random-vector controls provide a reference for interpreting activation-steering effects, but a single realization hides variation in that reference. We examine this variation and two separate implementation checks. In a released activation-addition (ActAdd) configuration on GPT-2-XL, we reproduce the steering arm over 57 prompts with a maximum absolute error of 1.74 × 10\ensuremath⁻\ensuremath⁶. At coefficient 10, 9 of 20 fresh random-control realizations have greater mean normalized Kullback–Leibler (KL) divergence than the steering arm, and 19 have lower means than the stored random realization. In a separate four-model study over 96 prompts, changing batch size or attention implementation changes the greedy next token for at most 5 and 6 prompts, respectively; an identical rerun changes none. Finally, an erroneous left-padding index in our own analysis pipeline, shared by the intervention and readout, overstates the maximum greedy-token change rate in 10 of 12 model-dose settings, with a median relative distortion of 46.8%. These findings support reporting realization-level variation, measuring configuration sensitivity, and validating intervention and readout indices. The results concern distributional change in the evaluated configurations, not semantic steering success.
%U https://aclanthology.org/2026.blackboxnlp-1.73/
%P 1264-1272
Markdown (Informal)
[Stress-Testing Random-Vector Controls for Activation Steering](https://aclanthology.org/2026.blackboxnlp-1.73/) (Mashayekhi et al., BlackboxNLP 2026)
ACL
- Behnood Mashayekhi, Parsa Hosseini, and Arian Amiramjadi. 2026. Stress-Testing Random-Vector Controls for Activation Steering. In Proceedings of the 9th BlackboxNLP Workshop: Analyzing and Interpreting Neural Networks for NLP, pages 1264–1272, Budapest, Hungary. Association for Computational Linguistics.