<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="tiwari-2025-extended">
    <titleInfo>
        <title>Extended Abstract: Probing-Guided Parameter-Efficient Fine-Tuning for Balancing Linguistic Adaptation and Safety in LLM-based Social Influence Systems</title>
    </titleInfo>
    <name type="personal">
        <namePart type="given">Manyana</namePart>
        <namePart type="family">Tiwari</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <originInfo>
        <dateIssued>2025-07</dateIssued>
    </originInfo>
    <typeOfResource>text</typeOfResource>
    <relatedItem type="host">
        <titleInfo>
            <title>Proceedings of the Third Workshop on Social Influence in Conversations (SICon 2025)</title>
        </titleInfo>
        <name type="personal">
            <namePart type="given">James</namePart>
            <namePart type="family">Hale</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <name type="personal">
            <namePart type="given">Brian</namePart>
            <namePart type="given">Deuksin</namePart>
            <namePart type="family">Kwon</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <name type="personal">
            <namePart type="given">Ritam</namePart>
            <namePart type="family">Dutt</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <originInfo>
            <publisher>Association for Computational Linguistics</publisher>
            <place>
                <placeTerm type="text">Vienna, Austria</placeTerm>
            </place>
        </originInfo>
        <genre authority="marcgt">conference publication</genre>
        <identifier type="isbn">979-8-89176-266-4</identifier>
    </relatedItem>
    <abstract>Designing effective LLMs for social influence (SI) tasks demands controlling linguistic output such that it adapts to context (such as user attributes, history etc.) while upholding ethical guardrails. Standard Parameter-Efficient Fine-Tuning (PEFT) methods like LoRA struggle to manage the trade-off between adaptive linguistic expression and safety and they optimize based on overall objectives without differentiating the functional roles of internal model components. Therefore, we introduce Probing-Guided PEFT (PG-PEFT), a novel fine-tuning strategy which utilizes interpretability probes to identify LLM components associated with context-driven linguistic variations versus those linked to safety violations (e.g., toxicity, bias). This functional map then guides LoRA updates, enabling more targeted control over the model’s linguistic output. We evaluate PG-PEFT on SI tasks (persuasion, negotiation) and linguistic adaptability with safety benchmarks against standard PEFT.</abstract>
    <identifier type="citekey">tiwari-2025-extended</identifier>
    <identifier type="doi">10.18653/v1/2025.sicon-1.12</identifier>
    <location>
        <url>https://aclanthology.org/2025.sicon-1.12/</url>
    </location>
    <part>
        <date>2025-07</date>
        <extent unit="page">
            <start>145</start>
            <end>147</end>
        </extent>
    </part>
</mods>
</modsCollection>

%0 Conference Proceedings
%T Extended Abstract: Probing-Guided Parameter-Efficient Fine-Tuning for Balancing Linguistic Adaptation and Safety in LLM-based Social Influence Systems
%A Tiwari, Manyana
%Y Hale, James
%Y Kwon, Brian Deuksin
%Y Dutt, Ritam
%S Proceedings of the Third Workshop on Social Influence in Conversations (SICon 2025)
%D 2025
%8 July
%I Association for Computational Linguistics
%C Vienna, Austria
%@ 979-8-89176-266-4
%F tiwari-2025-extended
%X Designing effective LLMs for social influence (SI) tasks demands controlling linguistic output such that it adapts to context (such as user attributes, history etc.) while upholding ethical guardrails. Standard Parameter-Efficient Fine-Tuning (PEFT) methods like LoRA struggle to manage the trade-off between adaptive linguistic expression and safety and they optimize based on overall objectives without differentiating the functional roles of internal model components. Therefore, we introduce Probing-Guided PEFT (PG-PEFT), a novel fine-tuning strategy which utilizes interpretability probes to identify LLM components associated with context-driven linguistic variations versus those linked to safety violations (e.g., toxicity, bias). This functional map then guides LoRA updates, enabling more targeted control over the model’s linguistic output. We evaluate PG-PEFT on SI tasks (persuasion, negotiation) and linguistic adaptability with safety benchmarks against standard PEFT.
%R 10.18653/v1/2025.sicon-1.12
%U https://aclanthology.org/2025.sicon-1.12/
%U https://doi.org/10.18653/v1/2025.sicon-1.12
%P 145-147

Markdown (Informal)

[Extended Abstract: Probing-Guided Parameter-Efficient Fine-Tuning for Balancing Linguistic Adaptation and Safety in LLM-based Social Influence Systems](https://aclanthology.org/2025.sicon-1.12/) (Tiwari, SICon 2025)

• Extended Abstract: Probing-Guided Parameter-Efficient Fine-Tuning for Balancing Linguistic Adaptation and Safety in LLM-based Social Influence Systems (Tiwari, SICon 2025)

ACL

• Manyana Tiwari. 2025. Extended Abstract: Probing-Guided Parameter-Efficient Fine-Tuning for Balancing Linguistic Adaptation and Safety in LLM-based Social Influence Systems. In Proceedings of the Third Workshop on Social Influence in Conversations (SICon 2025), pages 145–147, Vienna, Austria. Association for Computational Linguistics.