<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="yang-etal-2026-monolithic">
    <titleInfo>
        <title>From Monolithic Prompts to Scaffolded Execution for Memory Agents with Small Language Models</title>
    </titleInfo>
    <name type="personal">
        <namePart type="given">Seunghan</namePart>
        <namePart type="family">Yang</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Jihwan</namePart>
        <namePart type="family">Bang</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Sungha</namePart>
        <namePart type="family">Choi</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Juntae</namePart>
        <namePart type="family">Lee</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>Personal memory agents process sensitive user data, making on-device deployment with small language models (SLMs) attractive. However, existing memory-agent prompts are often monolithic specifications written for high-capacity LLMs, overloading a single inference with many requirements while leaving necessary intermediate operations implicit. This mismatch creates two bottlenecks for SLMs: prompt complexity and logical incompleteness. We introduce CAST (Compiling Agent Specifications into Scaffolded Tasks), a framework that converts a monolithic memory-agent prompt into a scaffolded execution procedure. CAST segments the original prompt into fixed, semantics-preserving main tasks, then synthesizes and places auxiliary scaffolding tasks that produce reusable intermediate artifacts; a trajectory-feedback step then adapts these scaffolds to the target SLM. At inference time, only the target SLM executes the compiled procedure. Across two memory-agent frameworks on LoCoMo and MemoryAgentBench, CAST outperforms monolithic prompting, prompt optimization, and reasoning-based baselines, while offering a stronger quality-latency trade-off than unconstrained reasoning. The results show that adapting memory agents to SLMs requires restructuring execution, not merely rewriting prompts.</abstract>
    <identifier type="citekey">yang-etal-2026-monolithic</identifier>
    <identifier type="doi">10.18653/2026.findings-emnlp.968</identifier>
    <location>
        <url>https://aclanthology.org/2026.findings-emnlp.968/</url>
    </location>
    <part>
        <date>2026-10</date>
        <extent unit="page">
            <start>20471</start>
            <end>20498</end>
        </extent>
    </part>
</mods>
</modsCollection>

%0 Conference Proceedings
%T From Monolithic Prompts to Scaffolded Execution for Memory Agents with Small Language Models
%A Yang, Seunghan
%A Bang, Jihwan
%A Choi, Sungha
%A Lee, Juntae
%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 yang-etal-2026-monolithic
%X Personal memory agents process sensitive user data, making on-device deployment with small language models (SLMs) attractive. However, existing memory-agent prompts are often monolithic specifications written for high-capacity LLMs, overloading a single inference with many requirements while leaving necessary intermediate operations implicit. This mismatch creates two bottlenecks for SLMs: prompt complexity and logical incompleteness. We introduce CAST (Compiling Agent Specifications into Scaffolded Tasks), a framework that converts a monolithic memory-agent prompt into a scaffolded execution procedure. CAST segments the original prompt into fixed, semantics-preserving main tasks, then synthesizes and places auxiliary scaffolding tasks that produce reusable intermediate artifacts; a trajectory-feedback step then adapts these scaffolds to the target SLM. At inference time, only the target SLM executes the compiled procedure. Across two memory-agent frameworks on LoCoMo and MemoryAgentBench, CAST outperforms monolithic prompting, prompt optimization, and reasoning-based baselines, while offering a stronger quality-latency trade-off than unconstrained reasoning. The results show that adapting memory agents to SLMs requires restructuring execution, not merely rewriting prompts.
%R 10.18653/2026.findings-emnlp.968
%U https://aclanthology.org/2026.findings-emnlp.968/
%U https://doi.org/10.18653/2026.findings-emnlp.968
%P 20471-20498

Markdown (Informal)

[From Monolithic Prompts to Scaffolded Execution for Memory Agents with Small Language Models](https://aclanthology.org/2026.findings-emnlp.968/) (Yang et al., Findings 2026)

• From Monolithic Prompts to Scaffolded Execution for Memory Agents with Small Language Models (Yang et al., Findings 2026)

ACL

• Seunghan Yang, Jihwan Bang, Sungha Choi, and Juntae Lee. 2026. From Monolithic Prompts to Scaffolded Execution for Memory Agents with Small Language Models. In Findings of the Association for Computational Linguistics: EMNLP 2026, pages 20471–20498, Budapest, Hungary. Association for Computational Linguistics.