@inproceedings{vinayaka-2026-llm,
title = "When Should {LLM} Agents Forget? Memory Filtering Under Query-Critical Resolver Errors",
author = "Vinayaka, Samarth",
editor = "Kamalloo, Ehsan and
Gontier, Nicolas and
Rozanov, Nikolai and
Hashemzadeh, Maryam and
Arora, Akhil",
booktitle = "Proceedings of the Second Workshop for {RE}search on Agent Language Models ({REALM} 2026)",
month = oct,
year = "2026",
address = "Budapest, Hungary",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2026.realm-1.16/",
pages = "212--225",
ISBN = "979-8-89176-485-9",
abstract = "LLM memory systems often filter facts classified as stale from the current prompt. We study this query-time context filtering, without deleting stored memories, in memory-conditioned question answering. Across three models on a synthetic benchmark, controlled label corruption moves hard-drop from gains of +0.186 to +0.230 at perfect resolution to losses of {\ensuremath{-}}0.176 to {\ensuremath{-}}0.181 at the lowest tested reliability. Paired targeted interventions show greater degradation from false-drop than false-keep; retaining evidence can still fail when incorrect labels steer generation. On one frozen 102/102 calibration/test query split, empirical policy selection using labeled calibration outcomes reaches 0.655 accuracy versus 0.641 for always hard-drop and 0.662 for test-informed oracle selection. A question-disjoint MemoryAgentBench replication is inconclusive: gain +0.03, 95{\%} CI [{\ensuremath{-}}0.01, +0.08]. These results establish a controlled sensitivity mechanism, not generalization to long-horizon agents. Global-accuracy crossovers require gold labels and are model- and task-specific diagnostics. Context filtering should be cost-sensitive and validated on matched labeled data; label-free retention is an accuracy fallback under the tested errors, subject to separate privacy and retention requirements."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="vinayaka-2026-llm">
<titleInfo>
<title>When Should LLM Agents Forget? Memory Filtering Under Query-Critical Resolver Errors</title>
</titleInfo>
<name type="personal">
<namePart type="given">Samarth</namePart>
<namePart type="family">Vinayaka</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 Second Workshop for REsearch on Agent Language Models (REALM 2026)</title>
</titleInfo>
<name type="personal">
<namePart type="given">Ehsan</namePart>
<namePart type="family">Kamalloo</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Nicolas</namePart>
<namePart type="family">Gontier</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Nikolai</namePart>
<namePart type="family">Rozanov</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Maryam</namePart>
<namePart type="family">Hashemzadeh</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Akhil</namePart>
<namePart type="family">Arora</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">979-8-89176-485-9</identifier>
</relatedItem>
<abstract>LLM memory systems often filter facts classified as stale from the current prompt. We study this query-time context filtering, without deleting stored memories, in memory-conditioned question answering. Across three models on a synthetic benchmark, controlled label corruption moves hard-drop from gains of +0.186 to +0.230 at perfect resolution to losses of \ensuremath-0.176 to \ensuremath-0.181 at the lowest tested reliability. Paired targeted interventions show greater degradation from false-drop than false-keep; retaining evidence can still fail when incorrect labels steer generation. On one frozen 102/102 calibration/test query split, empirical policy selection using labeled calibration outcomes reaches 0.655 accuracy versus 0.641 for always hard-drop and 0.662 for test-informed oracle selection. A question-disjoint MemoryAgentBench replication is inconclusive: gain +0.03, 95% CI [\ensuremath-0.01, +0.08]. These results establish a controlled sensitivity mechanism, not generalization to long-horizon agents. Global-accuracy crossovers require gold labels and are model- and task-specific diagnostics. Context filtering should be cost-sensitive and validated on matched labeled data; label-free retention is an accuracy fallback under the tested errors, subject to separate privacy and retention requirements.</abstract>
<identifier type="citekey">vinayaka-2026-llm</identifier>
<location>
<url>https://aclanthology.org/2026.realm-1.16/</url>
</location>
<part>
<date>2026-10</date>
<extent unit="page">
<start>212</start>
<end>225</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T When Should LLM Agents Forget? Memory Filtering Under Query-Critical Resolver Errors
%A Vinayaka, Samarth
%Y Kamalloo, Ehsan
%Y Gontier, Nicolas
%Y Rozanov, Nikolai
%Y Hashemzadeh, Maryam
%Y Arora, Akhil
%S Proceedings of the Second Workshop for REsearch on Agent Language Models (REALM 2026)
%D 2026
%8 October
%I Association for Computational Linguistics
%C Budapest, Hungary
%@ 979-8-89176-485-9
%F vinayaka-2026-llm
%X LLM memory systems often filter facts classified as stale from the current prompt. We study this query-time context filtering, without deleting stored memories, in memory-conditioned question answering. Across three models on a synthetic benchmark, controlled label corruption moves hard-drop from gains of +0.186 to +0.230 at perfect resolution to losses of \ensuremath-0.176 to \ensuremath-0.181 at the lowest tested reliability. Paired targeted interventions show greater degradation from false-drop than false-keep; retaining evidence can still fail when incorrect labels steer generation. On one frozen 102/102 calibration/test query split, empirical policy selection using labeled calibration outcomes reaches 0.655 accuracy versus 0.641 for always hard-drop and 0.662 for test-informed oracle selection. A question-disjoint MemoryAgentBench replication is inconclusive: gain +0.03, 95% CI [\ensuremath-0.01, +0.08]. These results establish a controlled sensitivity mechanism, not generalization to long-horizon agents. Global-accuracy crossovers require gold labels and are model- and task-specific diagnostics. Context filtering should be cost-sensitive and validated on matched labeled data; label-free retention is an accuracy fallback under the tested errors, subject to separate privacy and retention requirements.
%U https://aclanthology.org/2026.realm-1.16/
%P 212-225
Markdown (Informal)
[When Should LLM Agents Forget? Memory Filtering Under Query-Critical Resolver Errors](https://aclanthology.org/2026.realm-1.16/) (Vinayaka, REALM 2026)
ACL