@inproceedings{zhang-etal-2026-monolithic,
title = "From Monolithic Prioritization to Stratified Replay: Hierarchical Prioritized Experience Replay for Task-Oriented Dialogue Policy Learning",
author = "Zhang, Yixin and
Zhao, Yangyang and
Xu, Kai and
Xiang, Lingyun and
Song, Yun",
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.584/",
pages = "12336--12350",
ISBN = "979-8-89176-462-0",
abstract = "Replay sampling is central to off-policy task-oriented dialogue policy learning, as it determines which past interactions are revisited during policy updates. Existing priority-based replay methods often rely on a monolithic priority distribution to form minibatches, which entangles coverage across error magnitudes with emphasis on high-error transitions. Under sparse dialogue success and heavy-tailed temporal-difference (TD) errors, this design overexposes extreme residuals and suppresses informative medium-error transitions, a failure mode we call mid-magnitude starvation. To address this problem, we propose Hierarchical Prioritized Experience Replay (HPER), a two-level replay sampler that separates coverage from prioritization. HPER first samples TD-error quantile strata according to occupancy, ensuring that different error magnitudes enter training, and then applies local prioritized replay within the selected stratum to retain emphasis on large residuals. This coverage-first design restores intermediate learning signals without diluting high-error samples. Experiments on multiple dialogue datasets show that HPER improves replay efficiency and yields faster competence onset and higher terminal success than uniform replay and monolithic priority-based replay under matched budgets."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="zhang-etal-2026-monolithic">
<titleInfo>
<title>From Monolithic Prioritization to Stratified Replay: Hierarchical Prioritized Experience Replay for Task-Oriented Dialogue Policy Learning</title>
</titleInfo>
<name type="personal">
<namePart type="given">Yixin</namePart>
<namePart type="family">Zhang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yangyang</namePart>
<namePart type="family">Zhao</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Kai</namePart>
<namePart type="family">Xu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Lingyun</namePart>
<namePart type="family">Xiang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yun</namePart>
<namePart type="family">Song</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>Replay sampling is central to off-policy task-oriented dialogue policy learning, as it determines which past interactions are revisited during policy updates. Existing priority-based replay methods often rely on a monolithic priority distribution to form minibatches, which entangles coverage across error magnitudes with emphasis on high-error transitions. Under sparse dialogue success and heavy-tailed temporal-difference (TD) errors, this design overexposes extreme residuals and suppresses informative medium-error transitions, a failure mode we call mid-magnitude starvation. To address this problem, we propose Hierarchical Prioritized Experience Replay (HPER), a two-level replay sampler that separates coverage from prioritization. HPER first samples TD-error quantile strata according to occupancy, ensuring that different error magnitudes enter training, and then applies local prioritized replay within the selected stratum to retain emphasis on large residuals. This coverage-first design restores intermediate learning signals without diluting high-error samples. Experiments on multiple dialogue datasets show that HPER improves replay efficiency and yields faster competence onset and higher terminal success than uniform replay and monolithic priority-based replay under matched budgets.</abstract>
<identifier type="citekey">zhang-etal-2026-monolithic</identifier>
<location>
<url>https://aclanthology.org/2026.findings-emnlp.584/</url>
</location>
<part>
<date>2026-10</date>
<extent unit="page">
<start>12336</start>
<end>12350</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T From Monolithic Prioritization to Stratified Replay: Hierarchical Prioritized Experience Replay for Task-Oriented Dialogue Policy Learning
%A Zhang, Yixin
%A Zhao, Yangyang
%A Xu, Kai
%A Xiang, Lingyun
%A Song, Yun
%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 zhang-etal-2026-monolithic
%X Replay sampling is central to off-policy task-oriented dialogue policy learning, as it determines which past interactions are revisited during policy updates. Existing priority-based replay methods often rely on a monolithic priority distribution to form minibatches, which entangles coverage across error magnitudes with emphasis on high-error transitions. Under sparse dialogue success and heavy-tailed temporal-difference (TD) errors, this design overexposes extreme residuals and suppresses informative medium-error transitions, a failure mode we call mid-magnitude starvation. To address this problem, we propose Hierarchical Prioritized Experience Replay (HPER), a two-level replay sampler that separates coverage from prioritization. HPER first samples TD-error quantile strata according to occupancy, ensuring that different error magnitudes enter training, and then applies local prioritized replay within the selected stratum to retain emphasis on large residuals. This coverage-first design restores intermediate learning signals without diluting high-error samples. Experiments on multiple dialogue datasets show that HPER improves replay efficiency and yields faster competence onset and higher terminal success than uniform replay and monolithic priority-based replay under matched budgets.
%U https://aclanthology.org/2026.findings-emnlp.584/
%P 12336-12350
Markdown (Informal)
[From Monolithic Prioritization to Stratified Replay: Hierarchical Prioritized Experience Replay for Task-Oriented Dialogue Policy Learning](https://aclanthology.org/2026.findings-emnlp.584/) (Zhang et al., Findings 2026)
ACL