@inproceedings{tang-etal-2026-taylormlp,
title = "Why {T}aylor{MLP} Fails? Revisiting {LLM} Weight Protection Schemes with Adversarial Games",
author = "Tang, Bintao and
Yang, Xin and
Li, Letian and
Meng, Yu and
Lui, Kwok Wai and
Wang, Yuhao and
Ji, Zimo and
Zhang, Terry Jingchen and
Jiang, Wenyuan",
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.1378/",
pages = "29296--29310",
ISBN = "979-8-89176-462-0",
abstract = "Open-weight large language models create demand for weight protection schemes that let users run models locally while preserving developer control over valuable model weights. Existing schemes, however, are often justified by heuristic arguments and empirical checks rather than explicit adversarial goals, security definitions, or proofs. As a case study, we analyze a recent scheme previously published in *EMNLP 2024 Main Conference* and show that it is vulnerable to our efficient weight-extraction attack, recovering more than 99.95{\%} of protected parameters and producing reconstructed models with near-identical performance. Motivated by this failure, we formulate LLM weight protection schemes through game-based definitions, and we discuss how these notions relate. We encourage that future weight protection schemes should be designed with formal security goals, proof-oriented analysis, and offensive evaluation rather than relying on informal intuition alone."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="tang-etal-2026-taylormlp">
<titleInfo>
<title>Why TaylorMLP Fails? Revisiting LLM Weight Protection Schemes with Adversarial Games</title>
</titleInfo>
<name type="personal">
<namePart type="given">Bintao</namePart>
<namePart type="family">Tang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Xin</namePart>
<namePart type="family">Yang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Letian</namePart>
<namePart type="family">Li</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yu</namePart>
<namePart type="family">Meng</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Kwok</namePart>
<namePart type="given">Wai</namePart>
<namePart type="family">Lui</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yuhao</namePart>
<namePart type="family">Wang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Zimo</namePart>
<namePart type="family">Ji</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Terry</namePart>
<namePart type="given">Jingchen</namePart>
<namePart type="family">Zhang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Wenyuan</namePart>
<namePart type="family">Jiang</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>Open-weight large language models create demand for weight protection schemes that let users run models locally while preserving developer control over valuable model weights. Existing schemes, however, are often justified by heuristic arguments and empirical checks rather than explicit adversarial goals, security definitions, or proofs. As a case study, we analyze a recent scheme previously published in *EMNLP 2024 Main Conference* and show that it is vulnerable to our efficient weight-extraction attack, recovering more than 99.95% of protected parameters and producing reconstructed models with near-identical performance. Motivated by this failure, we formulate LLM weight protection schemes through game-based definitions, and we discuss how these notions relate. We encourage that future weight protection schemes should be designed with formal security goals, proof-oriented analysis, and offensive evaluation rather than relying on informal intuition alone.</abstract>
<identifier type="citekey">tang-etal-2026-taylormlp</identifier>
<location>
<url>https://aclanthology.org/2026.findings-emnlp.1378/</url>
</location>
<part>
<date>2026-10</date>
<extent unit="page">
<start>29296</start>
<end>29310</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Why TaylorMLP Fails? Revisiting LLM Weight Protection Schemes with Adversarial Games
%A Tang, Bintao
%A Yang, Xin
%A Li, Letian
%A Meng, Yu
%A Lui, Kwok Wai
%A Wang, Yuhao
%A Ji, Zimo
%A Zhang, Terry Jingchen
%A Jiang, Wenyuan
%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 tang-etal-2026-taylormlp
%X Open-weight large language models create demand for weight protection schemes that let users run models locally while preserving developer control over valuable model weights. Existing schemes, however, are often justified by heuristic arguments and empirical checks rather than explicit adversarial goals, security definitions, or proofs. As a case study, we analyze a recent scheme previously published in *EMNLP 2024 Main Conference* and show that it is vulnerable to our efficient weight-extraction attack, recovering more than 99.95% of protected parameters and producing reconstructed models with near-identical performance. Motivated by this failure, we formulate LLM weight protection schemes through game-based definitions, and we discuss how these notions relate. We encourage that future weight protection schemes should be designed with formal security goals, proof-oriented analysis, and offensive evaluation rather than relying on informal intuition alone.
%U https://aclanthology.org/2026.findings-emnlp.1378/
%P 29296-29310
Markdown (Informal)
[Why TaylorMLP Fails? Revisiting LLM Weight Protection Schemes with Adversarial Games](https://aclanthology.org/2026.findings-emnlp.1378/) (Tang et al., Findings 2026)
ACL
- Bintao Tang, Xin Yang, Letian Li, Yu Meng, Kwok Wai Lui, Yuhao Wang, Zimo Ji, Terry Jingchen Zhang, and Wenyuan Jiang. 2026. Why TaylorMLP Fails? Revisiting LLM Weight Protection Schemes with Adversarial Games. In Findings of the Association for Computational Linguistics: EMNLP 2026, pages 29296–29310, Budapest, Hungary. Association for Computational Linguistics.