<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="wang-etal-2026-hintsteer">
    <titleInfo>
        <title>HintSteer: Predicting Context Imprints for Inference-Time Steering</title>
    </titleInfo>
    <name type="personal">
        <namePart type="given">Xinda</namePart>
        <namePart type="family">Wang</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">ChenZhuo</namePart>
        <namePart type="family">Zhao</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">You</namePart>
        <namePart type="family">Hu</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Guotong</namePart>
        <namePart type="family">Li</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Chuang</namePart>
        <namePart type="family">Ou</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Yuhao</namePart>
        <namePart type="family">Zheng</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>Findings of the Association for Computational Linguistics: EMNLP 2026</title>
        </titleInfo>
        <name type="personal">
            <namePart type="given">Hung-Yi</namePart>
            <namePart type="family">Lee</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <name type="personal">
            <namePart type="given">Sunipa</namePart>
            <namePart type="family">Dev</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <name type="personal">
            <namePart type="given">Jessy</namePart>
            <namePart type="family">Li</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <name type="personal">
            <namePart type="given">Desmond</namePart>
            <namePart type="family">Elliott</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">979-8-89176-462-0</identifier>
    </relatedItem>
    <abstract>Large language models rely on inference-time context including rationales and instructions to improve task performance. Prior work attempts to internalize such benefits via parameter-efficient fine-tuning, and also studies inference-time representation interventions, but these approaches typically update the base model or learn fixed directions tied to global attributes. As a result, it remains unclear how to capture and reuse the internal effect of a helpful context that is specific to a given query. We take a representation-level view and define a context imprint as the hidden-state displacement between context-conditioned and query-only forward passes. We propose HintSteer, which trains a lightweight query-conditioned predictor to estimate the imprint from the query representation and injects it into a frozen model at inference time, approximating the missing context effect without explicitly providing context. Experiments on mathematical reasoning and writing demonstrate the effectiveness of imprint-based steering under no-context inference and indicate that both imprint magnitude and predictability influence the resulting gains.</abstract>
    <identifier type="citekey">wang-etal-2026-hintsteer</identifier>
    <location>
        <url>https://aclanthology.org/2026.findings-emnlp.656/</url>
    </location>
    <part>
        <date>2026-10</date>
        <extent unit="page">
            <start>13768</start>
            <end>13775</end>
        </extent>
    </part>
</mods>
</modsCollection>

%0 Conference Proceedings
%T HintSteer: Predicting Context Imprints for Inference-Time Steering
%A Wang, Xinda
%A Zhao, ChenZhuo
%A Hu, You
%A Li, Guotong
%A Ou, Chuang
%A Zheng, Yuhao
%Y Lee, Hung-Yi
%Y Dev, Sunipa
%Y Li, Jessy
%Y Elliott, Desmond
%S Findings of the Association for Computational Linguistics: EMNLP 2026
%D 2026
%8 October
%I Association for Computational Linguistics
%C Budapest, Hungary
%@ 979-8-89176-462-0
%F wang-etal-2026-hintsteer
%X Large language models rely on inference-time context including rationales and instructions to improve task performance. Prior work attempts to internalize such benefits via parameter-efficient fine-tuning, and also studies inference-time representation interventions, but these approaches typically update the base model or learn fixed directions tied to global attributes. As a result, it remains unclear how to capture and reuse the internal effect of a helpful context that is specific to a given query. We take a representation-level view and define a context imprint as the hidden-state displacement between context-conditioned and query-only forward passes. We propose HintSteer, which trains a lightweight query-conditioned predictor to estimate the imprint from the query representation and injects it into a frozen model at inference time, approximating the missing context effect without explicitly providing context. Experiments on mathematical reasoning and writing demonstrate the effectiveness of imprint-based steering under no-context inference and indicate that both imprint magnitude and predictability influence the resulting gains.
%U https://aclanthology.org/2026.findings-emnlp.656/
%P 13768-13775

Markdown (Informal)

[HintSteer: Predicting Context Imprints for Inference-Time Steering](https://aclanthology.org/2026.findings-emnlp.656/) (Wang et al., Findings 2026)

• HintSteer: Predicting Context Imprints for Inference-Time Steering (Wang et al., Findings 2026)

ACL

• Xinda Wang, ChenZhuo Zhao, You Hu, Guotong Li, Chuang Ou, and Yuhao Zheng. 2026. HintSteer: Predicting Context Imprints for Inference-Time Steering. In Findings of the Association for Computational Linguistics: EMNLP 2026, pages 13768–13775, Budapest, Hungary. Association for Computational Linguistics.