@inproceedings{huang-etal-2026-mem,
title = "Mem-{LM}: An {LM}-style Model for Reinforced Memory Selection",
author = "Huang, Yanhan and
Zhong, Yihao and
Ouyang, Wenwen and
Tian, Xing and
Yin, Zining and
Lu, Yiyi",
editor = "Atanasova, Pepa and
Haffari, Gholamreza and
Wang, Minghan",
booktitle = "Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing: Industry Track",
month = oct,
year = "2026",
address = "Budapest, Hungary",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2026.emnlp-industry.136/",
pages = "2290--2301",
ISBN = "979-8-89176-460-6",
abstract = "External memory banks can improve test-time reasoning in frozen large language models without updating backbone parameters. Yet similarity-based retrievers and pointwise selectors score memories independently, ignoring interactions and presentation order. We introduce , an LM-style model for reinforced memory selection that casts context construction as autoregressive index decoding. From a densely retrieved candidate pool, a lightweight Transformer encoder{--}decoder generates an ordered index sequence $[i_1,\ldots,i_k]$, modeling dependencies among memories while leaving the backbone frozen. This formulation supports standard language-model training. For sparse binary verifiable rewards, we further propose , a confidence-aware extension of group-relative policy optimization that promotes correct, low-confidence rollouts, penalizes confidently incorrect ones, and retains learning signals in reward-homogeneous groups. Across three mathematical reasoning and two reasoning-intensive question-answering benchmarks with three Qwen backbones, {~}achieves an average score of 84.9 on Qwen-plus, exceeding the strongest memory-augmented baseline by 2.3 points."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="huang-etal-2026-mem">
<titleInfo>
<title>Mem-LM: An LM-style Model for Reinforced Memory Selection</title>
</titleInfo>
<name type="personal">
<namePart type="given">Yanhan</namePart>
<namePart type="family">Huang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yihao</namePart>
<namePart type="family">Zhong</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Wenwen</namePart>
<namePart type="family">Ouyang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Xing</namePart>
<namePart type="family">Tian</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Zining</namePart>
<namePart type="family">Yin</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yiyi</namePart>
<namePart type="family">Lu</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: Industry Track</title>
</titleInfo>
<name type="personal">
<namePart type="given">Pepa</namePart>
<namePart type="family">Atanasova</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Gholamreza</namePart>
<namePart type="family">Haffari</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Minghan</namePart>
<namePart type="family">Wang</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">979-8-89176-460-6</identifier>
</relatedItem>
<abstract>External memory banks can improve test-time reasoning in frozen large language models without updating backbone parameters. Yet similarity-based retrievers and pointwise selectors score memories independently, ignoring interactions and presentation order. We introduce , an LM-style model for reinforced memory selection that casts context construction as autoregressive index decoding. From a densely retrieved candidate pool, a lightweight Transformer encoder–decoder generates an ordered index sequence [i₁,łdots,i_k], modeling dependencies among memories while leaving the backbone frozen. This formulation supports standard language-model training. For sparse binary verifiable rewards, we further propose , a confidence-aware extension of group-relative policy optimization that promotes correct, low-confidence rollouts, penalizes confidently incorrect ones, and retains learning signals in reward-homogeneous groups. Across three mathematical reasoning and two reasoning-intensive question-answering benchmarks with three Qwen backbones, achieves an average score of 84.9 on Qwen-plus, exceeding the strongest memory-augmented baseline by 2.3 points.</abstract>
<identifier type="citekey">huang-etal-2026-mem</identifier>
<location>
<url>https://aclanthology.org/2026.emnlp-industry.136/</url>
</location>
<part>
<date>2026-10</date>
<extent unit="page">
<start>2290</start>
<end>2301</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Mem-LM: An LM-style Model for Reinforced Memory Selection
%A Huang, Yanhan
%A Zhong, Yihao
%A Ouyang, Wenwen
%A Tian, Xing
%A Yin, Zining
%A Lu, Yiyi
%Y Atanasova, Pepa
%Y Haffari, Gholamreza
%Y Wang, Minghan
%S Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing: Industry Track
%D 2026
%8 October
%I Association for Computational Linguistics
%C Budapest, Hungary
%@ 979-8-89176-460-6
%F huang-etal-2026-mem
%X External memory banks can improve test-time reasoning in frozen large language models without updating backbone parameters. Yet similarity-based retrievers and pointwise selectors score memories independently, ignoring interactions and presentation order. We introduce , an LM-style model for reinforced memory selection that casts context construction as autoregressive index decoding. From a densely retrieved candidate pool, a lightweight Transformer encoder–decoder generates an ordered index sequence [i₁,łdots,i_k], modeling dependencies among memories while leaving the backbone frozen. This formulation supports standard language-model training. For sparse binary verifiable rewards, we further propose , a confidence-aware extension of group-relative policy optimization that promotes correct, low-confidence rollouts, penalizes confidently incorrect ones, and retains learning signals in reward-homogeneous groups. Across three mathematical reasoning and two reasoning-intensive question-answering benchmarks with three Qwen backbones, achieves an average score of 84.9 on Qwen-plus, exceeding the strongest memory-augmented baseline by 2.3 points.
%U https://aclanthology.org/2026.emnlp-industry.136/
%P 2290-2301
Markdown (Informal)
[Mem-LM: An LM-style Model for Reinforced Memory Selection](https://aclanthology.org/2026.emnlp-industry.136/) (Huang et al., EMNLP 2026)
ACL
- Yanhan Huang, Yihao Zhong, Wenwen Ouyang, Xing Tian, Zining Yin, and Yiyi Lu. 2026. Mem-LM: An LM-style Model for Reinforced Memory Selection. In Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing: Industry Track, pages 2290–2301, Budapest, Hungary. Association for Computational Linguistics.