@inproceedings{liu-etal-2026-helm,
title = "{HELM}: Steering Long-Horizon Agents with Learned Hierarchical Memory and Epistemic Governance",
author = "Liu, Jialong and
Shi, Yuling and
Li, Zuchao and
Wang, Ping",
editor = "Lee, Hung-Yi and
Dev, Sunipa and
Li, Jessy and
Elliott, Desmond",
booktitle = "Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing",
month = oct,
year = "2026",
address = "Budapest, Hungary",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2026.emnlp-main.721/",
pages = "16728--16753",
ISBN = "979-8-89176-459-0",
abstract = "Long-horizon LLM agents often fail not because relevant information is absent, but because semantically relevant memories are epistemically invalid: stale observations, failed tool traces, or claims contradicted by newer verified evidence are retrieved and treated as support. We introduce $\textbf{HELM} (\textbf{H}ierarchical \textbf{E}pistemic \textbf{L}earned \textbf{M}emory)$, a memory framework that makes this reliability problem explicit. HELM exposes memory through logged read/write/consolidate/prune operations and stores each item with provenance and epistemic state. Its three-tier $\textbf{Semantic-HNSW Nested Memory}$ (SHNM) links episodic tool traces to consolidated recalls and thematic indices, enabling coarse-to-fine retrieval that can be audited back to concrete evidence spans. $\emph{Epistemic governance}$ filters and re-ranks recalls by status, freshness, consistency, recency, and provenance, while a lightweight learned controller decides when memory operations are worth their cost. Across five long-horizon benchmarks, HELM improves success over strong memory baselines on GAIA, WebArena, Mind2Web, StreamBench, and LifelongAgentBench while reducing effective context tokens and epistemic errors. New scaling, training-free, and governance-only ablations show that the main gains persist with a 32B backbone and that epistemic governance alone explains a large share of the reliability improvement."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="liu-etal-2026-helm">
<titleInfo>
<title>HELM: Steering Long-Horizon Agents with Learned Hierarchical Memory and Epistemic Governance</title>
</titleInfo>
<name type="personal">
<namePart type="given">Jialong</namePart>
<namePart type="family">Liu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yuling</namePart>
<namePart type="family">Shi</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Zuchao</namePart>
<namePart type="family">Li</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Ping</namePart>
<namePart type="family">Wang</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>Long-horizon LLM agents often fail not because relevant information is absent, but because semantically relevant memories are epistemically invalid: stale observations, failed tool traces, or claims contradicted by newer verified evidence are retrieved and treated as support. We introduce HELM (Hierarchical Epistemic Learned Memory), a memory framework that makes this reliability problem explicit. HELM exposes memory through logged read/write/consolidate/prune operations and stores each item with provenance and epistemic state. Its three-tier Semantic-HNSW Nested Memory (SHNM) links episodic tool traces to consolidated recalls and thematic indices, enabling coarse-to-fine retrieval that can be audited back to concrete evidence spans. Epistemic governance filters and re-ranks recalls by status, freshness, consistency, recency, and provenance, while a lightweight learned controller decides when memory operations are worth their cost. Across five long-horizon benchmarks, HELM improves success over strong memory baselines on GAIA, WebArena, Mind2Web, StreamBench, and LifelongAgentBench while reducing effective context tokens and epistemic errors. New scaling, training-free, and governance-only ablations show that the main gains persist with a 32B backbone and that epistemic governance alone explains a large share of the reliability improvement.</abstract>
<identifier type="citekey">liu-etal-2026-helm</identifier>
<location>
<url>https://aclanthology.org/2026.emnlp-main.721/</url>
</location>
<part>
<date>2026-10</date>
<extent unit="page">
<start>16728</start>
<end>16753</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T HELM: Steering Long-Horizon Agents with Learned Hierarchical Memory and Epistemic Governance
%A Liu, Jialong
%A Shi, Yuling
%A Li, Zuchao
%A Wang, Ping
%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 liu-etal-2026-helm
%X Long-horizon LLM agents often fail not because relevant information is absent, but because semantically relevant memories are epistemically invalid: stale observations, failed tool traces, or claims contradicted by newer verified evidence are retrieved and treated as support. We introduce HELM (Hierarchical Epistemic Learned Memory), a memory framework that makes this reliability problem explicit. HELM exposes memory through logged read/write/consolidate/prune operations and stores each item with provenance and epistemic state. Its three-tier Semantic-HNSW Nested Memory (SHNM) links episodic tool traces to consolidated recalls and thematic indices, enabling coarse-to-fine retrieval that can be audited back to concrete evidence spans. Epistemic governance filters and re-ranks recalls by status, freshness, consistency, recency, and provenance, while a lightweight learned controller decides when memory operations are worth their cost. Across five long-horizon benchmarks, HELM improves success over strong memory baselines on GAIA, WebArena, Mind2Web, StreamBench, and LifelongAgentBench while reducing effective context tokens and epistemic errors. New scaling, training-free, and governance-only ablations show that the main gains persist with a 32B backbone and that epistemic governance alone explains a large share of the reliability improvement.
%U https://aclanthology.org/2026.emnlp-main.721/
%P 16728-16753
Markdown (Informal)
[HELM: Steering Long-Horizon Agents with Learned Hierarchical Memory and Epistemic Governance](https://aclanthology.org/2026.emnlp-main.721/) (Liu et al., EMNLP 2026)
ACL