<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="wang-etal-2026-learning-interaction">
    <titleInfo>
        <title>Learning Interaction-Centric Capability Representation for LLM Tool Calling</title>
    </titleInfo>
    <name type="personal">
        <namePart type="given">Hanyi</namePart>
        <namePart type="family">Wang</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Ruihan</namePart>
        <namePart type="family">Jin</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Jinyang</namePart>
        <namePart type="family">Wu</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Zhengqi</namePart>
        <namePart type="family">Wen</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Jianhua</namePart>
        <namePart type="family">Tao</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 2026 Conference on Empirical Methods in Natural Language Processing</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">979-8-89176-459-0</identifier>
    </relatedItem>
    <abstract>As large language models (LLMs) increasingly serve as the backbone of agentic systems with tool-use capabilities, predicting correctness, latency, and cost for a given tool-calling query is essential for efficient model evaluation, selection, and routing. Existing approaches predominantly adopt a semantics-first perspective, without capturing the complex interaction patterns inherent in tool-calling queries and without jointly accounting for cost and latency. We observe that query-model interaction matrices in tool-calling scenarios exhibit a low-rank structure, indicating that behavioral outcomes are governed by latent factors beyond surface semantics. We leverage this observation to propose a dual-pathway framework that decomposes interaction matrices into low-dimensional latent spaces for multi-objective prediction. The **Capability-First (CF)** pathway discovers latent structure via matrix factorization and then projects natural language queries into latent space. The **Joint Capability-Semantic (JCS)** pathway concurrently optimizes the text encoder and model embeddings, implicitly shaping query embeddings through capability-semantic interactions. Experiments on 75 LLMs demonstrate strong performance in predicting correctness, latency, and cost. Further experiments show that our method enables few-shot unseen-model generalization, parameter-efficient training, and interpretable latent representations.</abstract>
    <identifier type="citekey">wang-etal-2026-learning-interaction</identifier>
    <location>
        <url>https://aclanthology.org/2026.emnlp-main.1904/</url>
    </location>
    <part>
        <date>2026-10</date>
        <extent unit="page">
            <start>43394</start>
            <end>43409</end>
        </extent>
    </part>
</mods>
</modsCollection>

%0 Conference Proceedings
%T Learning Interaction-Centric Capability Representation for LLM Tool Calling
%A Wang, Hanyi
%A Jin, Ruihan
%A Wu, Jinyang
%A Wen, Zhengqi
%A Tao, Jianhua
%Y Lee, Hung-Yi
%Y Dev, Sunipa
%Y Li, Jessy
%Y Elliott, Desmond
%S Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing
%D 2026
%8 October
%I Association for Computational Linguistics
%C Budapest, Hungary
%@ 979-8-89176-459-0
%F wang-etal-2026-learning-interaction
%X As large language models (LLMs) increasingly serve as the backbone of agentic systems with tool-use capabilities, predicting correctness, latency, and cost for a given tool-calling query is essential for efficient model evaluation, selection, and routing. Existing approaches predominantly adopt a semantics-first perspective, without capturing the complex interaction patterns inherent in tool-calling queries and without jointly accounting for cost and latency. We observe that query-model interaction matrices in tool-calling scenarios exhibit a low-rank structure, indicating that behavioral outcomes are governed by latent factors beyond surface semantics. We leverage this observation to propose a dual-pathway framework that decomposes interaction matrices into low-dimensional latent spaces for multi-objective prediction. The **Capability-First (CF)** pathway discovers latent structure via matrix factorization and then projects natural language queries into latent space. The **Joint Capability-Semantic (JCS)** pathway concurrently optimizes the text encoder and model embeddings, implicitly shaping query embeddings through capability-semantic interactions. Experiments on 75 LLMs demonstrate strong performance in predicting correctness, latency, and cost. Further experiments show that our method enables few-shot unseen-model generalization, parameter-efficient training, and interpretable latent representations.
%U https://aclanthology.org/2026.emnlp-main.1904/
%P 43394-43409

Markdown (Informal)

[Learning Interaction-Centric Capability Representation for LLM Tool Calling](https://aclanthology.org/2026.emnlp-main.1904/) (Wang et al., EMNLP 2026)

• Learning Interaction-Centric Capability Representation for LLM Tool Calling (Wang et al., EMNLP 2026)

ACL

• Hanyi Wang, Ruihan Jin, Jinyang Wu, Zhengqi Wen, and Jianhua Tao. 2026. Learning Interaction-Centric Capability Representation for LLM Tool Calling. In Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing, pages 43394–43409, Budapest, Hungary. Association for Computational Linguistics.