<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="ghosh-etal-2026-scope">
    <titleInfo>
        <title>SCOPE: Stratum-Coordinated Optimization Pipeline for Efficient Agent Context</title>
    </titleInfo>
    <name type="personal">
        <namePart type="given">Pushpendu</namePart>
        <namePart type="family">Ghosh</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Aatir</namePart>
        <namePart type="family">Zaki</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Rudraraju</namePart>
        <namePart type="given">Sacin</namePart>
        <namePart type="family">Varma</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Stanley</namePart>
        <namePart type="family">Kanagaraj</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Anish</namePart>
        <namePart type="family">Nediyanchath</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: Industry Track</title>
        </titleInfo>
        <name type="personal">
            <namePart type="given">Pepa</namePart>
            <namePart type="family">Atanasova</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <name type="personal">
            <namePart type="given">Gholamreza</namePart>
            <namePart type="family">Haffari</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <name type="personal">
            <namePart type="given">Minghan</namePart>
            <namePart type="family">Wang</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">979-8-89176-460-6</identifier>
    </relatedItem>
    <abstract>LLM agents routinely spend orders of magnitude more tokens on context than on generation, re-sending interaction history, tool schemas, and system prompts at every turn. We introduce **SCOPE**, a unified framework for *context compilation* that compresses these three strata jointly under a fixed token budget while preserving agent behavior. SCOPE combines: 1. A learned *minimal sufficient context* reranker for memory, trained via leave-one-out supervision; 2. A plan-conditioned tool compressor that speculatively loads likely tools while retaining lightweight signatures for the rest; 3. A constraint-aware prompt compressor that preserves instructions. A lightweight planner coordinates these components via a contextual bandit, allocating budget across memory, tools, and prompt to maximize fidelity to full-context execution. On 3142 chat trajectories, SCOPE achieves **96.8%** decision fidelity at **4.46×** compression (60K tokens/turn), matching full-context task success within **0.9%** while using only **22%** of the tokens. Across operating regimes, SCOPE Pareto-dominates strong single-stratum baselines and naive compositions, with gains exceeding **11%** at matched cost. Our results show that optimizing context in isolation is insufficient: coordinated, budget-aware compression is key to scaling LLM agents without sacrificing reliability.</abstract>
    <identifier type="citekey">ghosh-etal-2026-scope</identifier>
    <location>
        <url>https://aclanthology.org/2026.emnlp-industry.129/</url>
    </location>
    <part>
        <date>2026-10</date>
        <extent unit="page">
            <start>2207</start>
            <end>2217</end>
        </extent>
    </part>
</mods>
</modsCollection>

%0 Conference Proceedings
%T SCOPE: Stratum-Coordinated Optimization Pipeline for Efficient Agent Context
%A Ghosh, Pushpendu
%A Zaki, Aatir
%A Varma, Rudraraju Sacin
%A Kanagaraj, Stanley
%A Nediyanchath, Anish
%Y Atanasova, Pepa
%Y Haffari, Gholamreza
%Y Wang, Minghan
%S Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing: Industry Track
%D 2026
%8 October
%I Association for Computational Linguistics
%C Budapest, Hungary
%@ 979-8-89176-460-6
%F ghosh-etal-2026-scope
%X LLM agents routinely spend orders of magnitude more tokens on context than on generation, re-sending interaction history, tool schemas, and system prompts at every turn. We introduce **SCOPE**, a unified framework for *context compilation* that compresses these three strata jointly under a fixed token budget while preserving agent behavior. SCOPE combines: 1. A learned *minimal sufficient context* reranker for memory, trained via leave-one-out supervision; 2. A plan-conditioned tool compressor that speculatively loads likely tools while retaining lightweight signatures for the rest; 3. A constraint-aware prompt compressor that preserves instructions. A lightweight planner coordinates these components via a contextual bandit, allocating budget across memory, tools, and prompt to maximize fidelity to full-context execution. On 3142 chat trajectories, SCOPE achieves **96.8%** decision fidelity at **4.46×** compression (60K tokens/turn), matching full-context task success within **0.9%** while using only **22%** of the tokens. Across operating regimes, SCOPE Pareto-dominates strong single-stratum baselines and naive compositions, with gains exceeding **11%** at matched cost. Our results show that optimizing context in isolation is insufficient: coordinated, budget-aware compression is key to scaling LLM agents without sacrificing reliability.
%U https://aclanthology.org/2026.emnlp-industry.129/
%P 2207-2217

Markdown (Informal)

[SCOPE: Stratum-Coordinated Optimization Pipeline for Efficient Agent Context](https://aclanthology.org/2026.emnlp-industry.129/) (Ghosh et al., EMNLP 2026)

• SCOPE: Stratum-Coordinated Optimization Pipeline for Efficient Agent Context (Ghosh et al., EMNLP 2026)

ACL

• Pushpendu Ghosh, Aatir Zaki, Rudraraju Sacin Varma, Stanley Kanagaraj, and Anish Nediyanchath. 2026. SCOPE: Stratum-Coordinated Optimization Pipeline for Efficient Agent Context. In Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing: Industry Track, pages 2207–2217, Budapest, Hungary. Association for Computational Linguistics.