Cover Image

Character as a Latent Variable in Large Language Models: A Mechanistic Account of Emergent Misalignment and Conditional Safety Failures

Fuente: arXiv

Saved in:

Bibliographic Details
Main Authors:	Su, Yanghao, Zhou, Wenbo, Zhang, Tianwei, Han, Qiu, Zhang, Weiming, Yu, Nenghai, Zhang, Jie
Format:	Preprint
Published:	2026
Subjects:	Computation and Language Artificial Intelligence Cryptography and Security
Online Access:	Acceder al recurso
Tags:	Add Tag No Tags, Be the first to tag this record!

Similar Items

State-Dependent Safety Failures in Multi-Turn Language Model Interaction
by: Li, Pengcheng, et al.
Published: (2026)

BURN: Backdoor Unlearning via Adversarial Boundary Analysis
by: Su, Yanghao, et al.
Published: (2025)

Model X-ray:Detecting Backdoored Models via Decision Boundary
by: Su, Yanghao, et al.
Published: (2024)

InferDPT: Privacy-Preserving Inference for Closed-box Large Language Model
by: Tong, Meng, et al.
Published: (2023)

AquaLoRA: Toward White-box Protection for Customized Stable Diffusion Models via Watermark LoRA
by: Feng, Weitao, et al.
Published: (2024)

SQL Injection Jailbreak: A Structural Disaster of Large Language Models
by: Zhao, Jiawei, et al.
Published: (2024)

SafeGuider: Robust and Practical Content Safety Control for Text-to-Image Models
by: Qi, Peigui, et al.
Published: (2025)

PoseGuard: Pose-Guided Generation with Safety Guardrails
by: Wang, Kongxin, et al.
Published: (2025)

Turning Your Strength into Watermark: Watermarking Large Language Model via Knowledge Injection
by: Li, Shuai, et al.
Published: (2023)

EditMark: Watermarking Large Language Models based on Model Editing
by: Li, Shuai, et al.
Published: (2025)

Silent Guardian: Protecting Text from Malicious Exploitation by Large Language Models
by: Zhao, Jiawei, et al.
Published: (2023)

Performance-lossless Black-box Model Watermarking
by: Zhao, Na, et al.
Published: (2023)

When Search Goes Wrong: Red-Teaming Web-Augmented Large Language Models
by: Ou, Haoran, et al.
Published: (2025)

ObfusBFA: A Holistic Approach to Safeguarding DNNs from Different Types of Bit-Flip Attacks
by: Yan, Xiaobei, et al.
Published: (2025)

On the Vulnerability of Text Sanitization
by: Tong, Meng, et al.
Published: (2024)

PSRT: Accelerating LRM-based Guard Models via Prefilled Safe Reasoning Traces
by: Zhao, Jiawei, et al.
Published: (2025)

Provably Secure Agent Guardrail
by: Wu, Benlong, et al.
Published: (2026)

A high-capacity linguistic steganography based on entropy-driven rank-token mapping
by: Jiang, Jun, et al.
Published: (2025)

STEAD: Robust Provably Secure Linguistic Steganography with Diffusion Language Model
by: Qi, Yuang, et al.
Published: (2026)

BitHydra: Towards Bit-flip Inference Cost Attack against Large Language Models
by: Yan, Xiaobei, et al.
Published: (2025)

IRCopilot: Automated Incident Response with Large Language Models
by: Lin, Xihuan, et al.
Published: (2025)

Provably Secure Public-Key Steganography Based on Admissible Encoding
by: Zhang, Xin, et al.
Published: (2025)

An Empirical Study on the Effectiveness of Large Language Models for Binary Code Understanding
by: Shang, Xiuwei, et al.
Published: (2025)

GIFDL: Generated Image Fluctuation Distortion Learning for Enhancing Steganographic Security
by: Wang, Xiangkun, et al.
Published: (2025)

FoC: Figure out the Cryptographic Functions in Stripped Binaries with LLMs
by: Shang, Xiuwei, et al.
Published: (2024)

How Far Have We Gone in Binary Code Understanding Using Large Language Models
by: Shang, Xiuwei, et al.
Published: (2024)

PRIVMARK: Private Large Language Models Watermarking with MPC
by: Fargues, Thomas, et al.
Published: (2025)

An Engorgio Prompt Makes Large Language Model Babble on
by: Dong, Jianshuo, et al.
Published: (2024)

Provably Secure Disambiguating Neural Linguistic Steganography
by: Qi, Yuang, et al.
Published: (2024)

Eliciting and Analyzing Emergent Misalignment in State-of-the-Art Large Language Models
by: Panpatil, Siddhant, et al.
Published: (2025)

Beyond the Edge of Function: Unraveling the Patterns of Type Recovery in Binary Code
by: Li, Gangyang, et al.
Published: (2025)

BadLingual: A Novel Lingual-Backdoor Attack against Large Language Models
by: Wang, Zihan, et al.
Published: (2025)

Towards Action Hijacking of Large Language Model-based Agent
by: Zhang, Yuyang, et al.
Published: (2024)

Exploiting Vulnerabilities in Speech Translation Systems through Targeted Adversarial Attacks
by: Liu, Chang, et al.
Published: (2025)

SafeBench: A Safety Evaluation Framework for Multimodal Large Language Models
by: Ying, Zonghao, et al.
Published: (2024)

Gaussian Shading: Provable Performance-Lossless Image Watermarking for Diffusion Models
by: Yang, Zijin, et al.
Published: (2024)

Robust-Wide: Robust Watermarking against Instruction-driven Image Editing
by: Hu, Runyi, et al.
Published: (2024)

VertMark: A Unified Training-Free Robust Watermarking Framework for Vertical Domain Pre-trained Language Models
by: Kong, Cong, et al.
Published: (2026)

AISA: Awakening Intrinsic Safety Awareness in Large Language Models against Jailbreak Attacks
by: Song, Weiming, et al.
Published: (2026)

Safeguarding Large Language Models: A Survey
by: Dong, Yi, et al.
Published: (2024)