@inproceedings{cao-etal-2026-reservegate,
title = "{R}eserve{G}ate: Reserve-Aware Evidence Scheduling for Long-Horizon Language Agents",
author = "Cao, Shuang and
Li, Rui and
Liu, Louise Ruihua and
Duprey, Alexandre",
editor = "Lee, Hung-Yi and
Dev, Sunipa and
Li, Jessy and
Elliott, Desmond",
booktitle = "Findings of the {A}ssociation for {C}omputational {L}inguistics: {EMNLP} 2026",
month = oct,
year = "2026",
address = "Budapest, Hungary",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2026.findings-emnlp.50/",
pages = "998--1017",
ISBN = "979-8-89176-462-0",
abstract = "Long-horizon language agents must decide not only which evidence to use, but also when to pay for it. Retrieval and verification can improve a local decision while exhausting token and tool-time budgets needed later in the episode. We introduce ReserveGate, a one-step evidence controller that combines an explicit uncertainty sketch, an off-policy action-value model, and component-wise feasibility masks over the remaining resources. Unlike a fixed reserve heuristic, ReserveGate adapts its cost penalty when spending runs ahead of episode progress and gradually releases a protected late-stage reserve. Across three benchmark-derived agent environments, ReserveGate improves tight-budget success over the strongest matched controller by 2.6{--}3.8 percentage points, with gains of 4.6{--}6.2 points on the longest episode quartile. A latency-constrained variant reduces service-level incidents to 1.2{--}1.7{\%} across environments while losing at most 0.8 points in success. On LoCoMo-Episodic, cross-model, cross-retriever, rate-limiter, and task-agnostic-sketch comparisons show that the effect is not explained by a single backbone, retrieval stack, or tuned spending rule."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="cao-etal-2026-reservegate">
<titleInfo>
<title>ReserveGate: Reserve-Aware Evidence Scheduling for Long-Horizon Language Agents</title>
</titleInfo>
<name type="personal">
<namePart type="given">Shuang</namePart>
<namePart type="family">Cao</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Rui</namePart>
<namePart type="family">Li</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Louise</namePart>
<namePart type="given">Ruihua</namePart>
<namePart type="family">Liu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Alexandre</namePart>
<namePart type="family">Duprey</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>Long-horizon language agents must decide not only which evidence to use, but also when to pay for it. Retrieval and verification can improve a local decision while exhausting token and tool-time budgets needed later in the episode. We introduce ReserveGate, a one-step evidence controller that combines an explicit uncertainty sketch, an off-policy action-value model, and component-wise feasibility masks over the remaining resources. Unlike a fixed reserve heuristic, ReserveGate adapts its cost penalty when spending runs ahead of episode progress and gradually releases a protected late-stage reserve. Across three benchmark-derived agent environments, ReserveGate improves tight-budget success over the strongest matched controller by 2.6–3.8 percentage points, with gains of 4.6–6.2 points on the longest episode quartile. A latency-constrained variant reduces service-level incidents to 1.2–1.7% across environments while losing at most 0.8 points in success. On LoCoMo-Episodic, cross-model, cross-retriever, rate-limiter, and task-agnostic-sketch comparisons show that the effect is not explained by a single backbone, retrieval stack, or tuned spending rule.</abstract>
<identifier type="citekey">cao-etal-2026-reservegate</identifier>
<location>
<url>https://aclanthology.org/2026.findings-emnlp.50/</url>
</location>
<part>
<date>2026-10</date>
<extent unit="page">
<start>998</start>
<end>1017</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T ReserveGate: Reserve-Aware Evidence Scheduling for Long-Horizon Language Agents
%A Cao, Shuang
%A Li, Rui
%A Liu, Louise Ruihua
%A Duprey, Alexandre
%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 cao-etal-2026-reservegate
%X Long-horizon language agents must decide not only which evidence to use, but also when to pay for it. Retrieval and verification can improve a local decision while exhausting token and tool-time budgets needed later in the episode. We introduce ReserveGate, a one-step evidence controller that combines an explicit uncertainty sketch, an off-policy action-value model, and component-wise feasibility masks over the remaining resources. Unlike a fixed reserve heuristic, ReserveGate adapts its cost penalty when spending runs ahead of episode progress and gradually releases a protected late-stage reserve. Across three benchmark-derived agent environments, ReserveGate improves tight-budget success over the strongest matched controller by 2.6–3.8 percentage points, with gains of 4.6–6.2 points on the longest episode quartile. A latency-constrained variant reduces service-level incidents to 1.2–1.7% across environments while losing at most 0.8 points in success. On LoCoMo-Episodic, cross-model, cross-retriever, rate-limiter, and task-agnostic-sketch comparisons show that the effect is not explained by a single backbone, retrieval stack, or tuned spending rule.
%U https://aclanthology.org/2026.findings-emnlp.50/
%P 998-1017
Markdown (Informal)
[ReserveGate: Reserve-Aware Evidence Scheduling for Long-Horizon Language Agents](https://aclanthology.org/2026.findings-emnlp.50/) (Cao et al., Findings 2026)
ACL