@inproceedings{zhang-etal-2026-bidirectional-lms,
title = "Bidirectional {LM}s are Better Knowledge Memorizers? A Benchmark for Real-world Knowledge Injection",
author = "Zhang, Yuwei and
Yu, Wenhao and
Feng, Shangbin and
Zhu, Yifan and
Peng, Letian and
Srinivasa, Jayanth and
Liu, Gaowen and
Shang, Jingbo",
editor = "Liakata, Maria and
Moreira, Viviane P. and
Zhang, Jiajun and
Jurgens, David",
booktitle = "Proceedings of the 64th Annual Meeting of the {A}ssociation for {C}omputational {L}inguistics (Volume 1: Long Papers)",
month = jul,
year = "2026",
address = "San Diego, California, United States",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2026.acl-long.771/",
pages = "16964--16980",
ISBN = "979-8-89176-390-6",
abstract = "Despite significant advances in large language models (LLMs), their knowledge memorization capabilities remain underexplored, due to the lack of standardized and high-quality testing grounds. In this paper, we introduce a novel, real-world and large-scale knowledge injection benchmark that continuously evolves without human intervention. Specifically, we propose WikiDYK, which leverages recently-added and expert-curated facts from Wikipedia{'}s ``Did You Know...'' entries. Each entry is converted into multiple question{--}answer pairs spanning diverse task formats from easy cloze prompts to complex multi-hop questions. WikiDYK currently contains 12,290 facts and 77,180 questions, and its design allows for seamless extension with future updates from Wikipedia editors. Through extensive experiments using continued pre-training, we reveal a surprising insight: despite their prevalence in modern LLMs, Causal Language Models (CLMs) demonstrate significantly weaker knowledge memorization capabilities compared to Bidirectional Language Models (BiLMs), exhibiting a 23{\%} lower accuracy in terms of reliability. To compensate for the smaller scales of current BiLMs, we introduce a modular collaborative framework utilizing ensembles of BiLMs as external knowledge repositories to integrate with LLMs. Experiment shows that this framework further improves the reliability accuracy by up to 29.1{\%}. Code: \url{https://github.com/zhang-yu-wei/WikiDYK}."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="zhang-etal-2026-bidirectional-lms">
<titleInfo>
<title>Bidirectional LMs are Better Knowledge Memorizers? A Benchmark for Real-world Knowledge Injection</title>
</titleInfo>
<name type="personal">
<namePart type="given">Yuwei</namePart>
<namePart type="family">Zhang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Wenhao</namePart>
<namePart type="family">Yu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Shangbin</namePart>
<namePart type="family">Feng</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yifan</namePart>
<namePart type="family">Zhu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Letian</namePart>
<namePart type="family">Peng</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Jayanth</namePart>
<namePart type="family">Srinivasa</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Gaowen</namePart>
<namePart type="family">Liu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Jingbo</namePart>
<namePart type="family">Shang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<originInfo>
<dateIssued>2026-07</dateIssued>
</originInfo>
<typeOfResource>text</typeOfResource>
<relatedItem type="host">
<titleInfo>
<title>Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)</title>
</titleInfo>
<name type="personal">
<namePart type="given">Maria</namePart>
<namePart type="family">Liakata</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Viviane</namePart>
<namePart type="given">P</namePart>
<namePart type="family">Moreira</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Jiajun</namePart>
<namePart type="family">Zhang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">David</namePart>
<namePart type="family">Jurgens</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<originInfo>
<publisher>Association for Computational Linguistics</publisher>
<place>
<placeTerm type="text">San Diego, California, United States</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
<identifier type="isbn">979-8-89176-390-6</identifier>
</relatedItem>
<abstract>Despite significant advances in large language models (LLMs), their knowledge memorization capabilities remain underexplored, due to the lack of standardized and high-quality testing grounds. In this paper, we introduce a novel, real-world and large-scale knowledge injection benchmark that continuously evolves without human intervention. Specifically, we propose WikiDYK, which leverages recently-added and expert-curated facts from Wikipedia’s “Did You Know...” entries. Each entry is converted into multiple question–answer pairs spanning diverse task formats from easy cloze prompts to complex multi-hop questions. WikiDYK currently contains 12,290 facts and 77,180 questions, and its design allows for seamless extension with future updates from Wikipedia editors. Through extensive experiments using continued pre-training, we reveal a surprising insight: despite their prevalence in modern LLMs, Causal Language Models (CLMs) demonstrate significantly weaker knowledge memorization capabilities compared to Bidirectional Language Models (BiLMs), exhibiting a 23% lower accuracy in terms of reliability. To compensate for the smaller scales of current BiLMs, we introduce a modular collaborative framework utilizing ensembles of BiLMs as external knowledge repositories to integrate with LLMs. Experiment shows that this framework further improves the reliability accuracy by up to 29.1%. Code: https://github.com/zhang-yu-wei/WikiDYK.</abstract>
<identifier type="citekey">zhang-etal-2026-bidirectional-lms</identifier>
<location>
<url>https://aclanthology.org/2026.acl-long.771/</url>
</location>
<part>
<date>2026-07</date>
<extent unit="page">
<start>16964</start>
<end>16980</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Bidirectional LMs are Better Knowledge Memorizers? A Benchmark for Real-world Knowledge Injection
%A Zhang, Yuwei
%A Yu, Wenhao
%A Feng, Shangbin
%A Zhu, Yifan
%A Peng, Letian
%A Srinivasa, Jayanth
%A Liu, Gaowen
%A Shang, Jingbo
%Y Liakata, Maria
%Y Moreira, Viviane P.
%Y Zhang, Jiajun
%Y Jurgens, David
%S Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)
%D 2026
%8 July
%I Association for Computational Linguistics
%C San Diego, California, United States
%@ 979-8-89176-390-6
%F zhang-etal-2026-bidirectional-lms
%X Despite significant advances in large language models (LLMs), their knowledge memorization capabilities remain underexplored, due to the lack of standardized and high-quality testing grounds. In this paper, we introduce a novel, real-world and large-scale knowledge injection benchmark that continuously evolves without human intervention. Specifically, we propose WikiDYK, which leverages recently-added and expert-curated facts from Wikipedia’s “Did You Know...” entries. Each entry is converted into multiple question–answer pairs spanning diverse task formats from easy cloze prompts to complex multi-hop questions. WikiDYK currently contains 12,290 facts and 77,180 questions, and its design allows for seamless extension with future updates from Wikipedia editors. Through extensive experiments using continued pre-training, we reveal a surprising insight: despite their prevalence in modern LLMs, Causal Language Models (CLMs) demonstrate significantly weaker knowledge memorization capabilities compared to Bidirectional Language Models (BiLMs), exhibiting a 23% lower accuracy in terms of reliability. To compensate for the smaller scales of current BiLMs, we introduce a modular collaborative framework utilizing ensembles of BiLMs as external knowledge repositories to integrate with LLMs. Experiment shows that this framework further improves the reliability accuracy by up to 29.1%. Code: https://github.com/zhang-yu-wei/WikiDYK.
%U https://aclanthology.org/2026.acl-long.771/
%P 16964-16980
Markdown (Informal)
[Bidirectional LMs are Better Knowledge Memorizers? A Benchmark for Real-world Knowledge Injection](https://aclanthology.org/2026.acl-long.771/) (Zhang et al., ACL 2026)
ACL
- Yuwei Zhang, Wenhao Yu, Shangbin Feng, Yifan Zhu, Letian Peng, Jayanth Srinivasa, Gaowen Liu, and Jingbo Shang. 2026. Bidirectional LMs are Better Knowledge Memorizers? A Benchmark for Real-world Knowledge Injection. In Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 16964–16980, San Diego, California, United States. Association for Computational Linguistics.