Cover Image

Token by Token, Compromised: Backdoor Vulnerabilities in Unified Autoregressive Models

Fuente: arXiv

Saved in:

Bibliographic Details
Main Authors:	Braun, Tobias, Grebe, Jonas Henry, Shakibania, Hossein, Rohrbach, Anna, Rohrbach, Marcus
Format:	Preprint
Published:	2026
Subjects:	Cryptography and Security Artificial Intelligence
Online Access:	Acceder al recurso
Tags:	Add Tag No Tags, Be the first to tag this record!

Similar Items

Erased but Not Forgotten: How Backdoors Compromise Concept Erasure
by: Braun, Tobias, et al.
Published: (2025)

Geometric Erasure by Contrastive Velocity Matching in Rectified Flows
by: Grebe, Jonas Henry, et al.
Published: (2026)

Backdoor Token Unlearning: Exposing and Defending Backdoors in Pretrained Language Models
by: Jiang, Peihai, et al.
Published: (2025)

Compromising Embodied Agents with Contextual Backdoor Attacks
by: Liu, Aishan, et al.
Published: (2024)

Exploring Backdoor Vulnerabilities of Chat Models
by: Hao, Yunzhuo, et al.
Published: (2024)

Membership Inference Attacks on Tokenizers of Large Language Models
by: Tong, Meng, et al.
Published: (2025)

Selection-Based Vulnerabilities: Clean-Label Backdoor Attacks in Active Learning
by: Zhi, Yuhan, et al.
Published: (2025)

Token-Efficient Prompt Injection Attack: Provoking Cessation in LLM Reasoning via Adaptive Token Compression
by: Cui, Yu, et al.
Published: (2025)

Backdoor Attribution: Elucidating and Controlling Backdoor in Language Models
by: Yu, Miao, et al.
Published: (2025)

Reliable Model Watermarking: Defending Against Theft without Compromising on Evasion
by: Zhu, Hongyu, et al.
Published: (2024)

Universal Vulnerabilities in Large Language Models: Backdoor Attacks for In-context Learning
by: Zhao, Shuai, et al.
Published: (2024)

Instructions as Backdoors: Backdoor Vulnerabilities of Instruction Tuning for Large Language Models
by: Xu, Jiashu, et al.
Published: (2023)

BadJudge: Backdoor Vulnerabilities of LLM-as-a-Judge
by: Tong, Terry, et al.
Published: (2025)

MTRE: Multi-Token Reliability Estimation for Hallucination Detection in VLMs
by: Zollicoffer, Geigh, et al.
Published: (2025)

Exploiting the Vulnerability of Large Language Models via Defense-Aware Architectural Backdoor
by: Miah, Abdullah Arafat, et al.
Published: (2024)

GUARD-SLM: Token Activation-Based Defense Against Jailbreak Attacks for Small Language Models
by: Mia, Md Jueal, et al.
Published: (2026)

Unified Neural Backdoor Removal with Only Few Clean Samples through Unlearning and Relearning
by: Min, Nay Myat, et al.
Published: (2024)

Backdoor Sentinel: Detecting and Detoxifying Backdoors in Diffusion Models via Temporal Noise Consistency
by: Wang, Bingzheng, et al.
Published: (2026)

When Backdoors Speak: Understanding LLM Backdoor Attacks Through Model-Generated Explanations
by: Ge, Huaizhi, et al.
Published: (2024)

Understanding Secret Leakage Risks in Code LLMs: A Tokenization Perspective
by: Chen, Meifang, et al.
Published: (2026)

Privacy Guard & Token Parsimony by Prompt and Context Handling and LLM Routing
by: Langiu, Alessio
Published: (2026)

AI-Governed Agent Architecture for Web-Trustworthy Tokenization of Alternative Assets
by: Borjigin, Ailiya, et al.
Published: (2025)

UTF:Undertrained Tokens as Fingerprints A Novel Approach to LLM Identification
by: Cai, Jiacheng, et al.
Published: (2024)

TokenMark: A Modality-Agnostic Watermark for Pre-trained Transformers
by: Xu, Hengyuan, et al.
Published: (2024)

Less Is More -- Until It Breaks: Security Pitfalls of Vision Token Compression in Large Vision-Language Models
by: Zhang, Xiaomei, et al.
Published: (2026)

Exploiting Layer-Specific Vulnerabilities to Backdoor Attack in Federated Learning
by: Foroughi, Mohammad Hadi, et al.
Published: (2026)

Backdooring Bias in Large Language Models
by: Das, Anudeep, et al.
Published: (2026)

Lightweight and Fast Backdoor Model Detection
by: Yu, Yinbo, et al.
Published: (2026)

DeBackdoor: A Deductive Framework for Detecting Backdoor Attacks on Deep Models with Limited Data
by: Popovic, Dorde, et al.
Published: (2025)

BackdoorMBTI: A Backdoor Learning Multimodal Benchmark Tool Kit for Backdoor Defense Evaluation
by: Yu, Haiyang, et al.
Published: (2024)

MetaBreak: Jailbreaking Online LLM Services via Special Token Manipulation
by: Zhu, Wentian, et al.
Published: (2025)

Auditing Pay-Per-Token in Large Language Models
by: Velasco, Ander Artola, et al.
Published: (2025)

TERD: A Unified Framework for Safeguarding Diffusion Models Against Backdoors
by: Mo, Yichuan, et al.
Published: (2024)

Towards Backdoor Stealthiness in Model Parameter Space
by: Xu, Xiaoyun, et al.
Published: (2025)

Are My Optimized Prompts Compromised? Exploring Vulnerabilities of LLM-based Optimizers
by: Zhao, Andrew, et al.
Published: (2025)

Backdoor4Good: Benchmarking Beneficial Uses of Backdoors in LLMs
by: Li, Yige, et al.
Published: (2026)

Backdoors in RLVR: Jailbreak Backdoors in LLMs From Verifiable Reward
by: Guo, Weiyang, et al.
Published: (2026)

AutoBackdoor: Automating Backdoor Attacks via LLM Agents
by: Li, Yige, et al.
Published: (2025)

Safety Alignment Should Be Made More Than Just a Few Tokens Deep
by: Qi, Xiangyu, et al.
Published: (2024)

Kill Two Birds with One Stone! Trajectory enabled Unified Online Detection of Adversarial Examples and Backdoor Attacks
by: Fu, Anmin, et al.
Published: (2025)