Copertina

Latent Fusion Jailbreak: Blending Harmful and Harmless Representations to Elicit Unsafe LLM Outputs

Fuente: arXiv

Salvato in:

Dettagli Bibliografici
Autori principali:	Xing, Wenpeng, Li, Mohan, Hu, Chunqiang, Xu, Haitao, Zhang, Ningyu, Lin, Bo, Han, Meng
Natura:	Preprint
Pubblicazione:	2025
Soggetti:	Computation and Language Artificial Intelligence Cryptography and Security
Accesso online:	Acceder al recurso
Tags:	Aggiungi Tag Nessun Tag, puoi essere il primo ad aggiungerne!!

Documenti analoghi

PREE: Towards Harmless and Adaptive Fingerprint Editing in Large Language Models via Knowledge Prefix Enhancement
di: Yue, Xubin, et al.
Pubblicazione: (2025)

EverTracer: Hunting Stolen Large Language Models via Stealthy and Robust Probabilistic Fingerprint
di: Xu, Zhenhua, et al.
Pubblicazione: (2025)

Understanding LLM Behavior When Encountering User-Supplied Harmful Content in Harmless Tasks
di: Chu, Junjie, et al.
Pubblicazione: (2026)

Towards Robust and Secure Embodied AI: A Survey on Vulnerabilities and Attacks
di: Xing, Wenpeng, et al.
Pubblicazione: (2025)

When Benign Inputs Lead to Severe Harms: Eliciting Unsafe Unintended Behaviors of Computer-Use Agents
di: Jones, Jaylen, et al.
Pubblicazione: (2026)

NEXUS: Network Exploration for eXploiting Unsafe Sequences in Multi-Turn LLM Jailbreaks
di: Asl, Javad Rafiei, et al.
Pubblicazione: (2025)

ForgetMark: Stealthy Fingerprint Embedding via Targeted Unlearning in Language Models
di: Xu, Zhenhua, et al.
Pubblicazione: (2026)

HarmChip: Evaluating Hardware Security Centric LLM Safety via Jailbreak Benchmarking
di: Wang, Zeng, et al.
Pubblicazione: (2026)

Harmless Yet Harmful: Neutral Prompting Attacks for Stealthy Hallucination Steering in Agent Skills
di: Hsu, Chia-Yi, et al.
Pubblicazione: (2026)

Eliciting Harmful Capabilities by Fine-Tuning On Safeguarded Outputs
di: Kaunismaa, Jackson, et al.
Pubblicazione: (2026)

LLMs can be Dangerous Reasoners: Analyzing-based Jailbreak Attack on Large Language Models
di: Lin, Shi, et al.
Pubblicazione: (2024)

Differentially Private Distance Query with Asymmetric Noise
di: Sheng, Weihong, et al.
Pubblicazione: (2025)

Local Distance Query with Differential Privacy
di: Sheng, Weihong, et al.
Pubblicazione: (2025)

JailbreakLens: Interpreting Jailbreak Mechanism in the Lens of Representation and Circuit
di: He, Zeqing, et al.
Pubblicazione: (2024)

MEraser: An Effective Fingerprint Erasure Approach for Large Language Models
di: Zhang, Jingxuan, et al.
Pubblicazione: (2025)

Improving LLM Outputs Against Jailbreak Attacks with Expert Model Integration
di: Tsmindashvili, Tatia, et al.
Pubblicazione: (2025)

GenTel-Safe: A Unified Benchmark and Shielding Framework for Defending Against Prompt Injection Attacks
di: Li, Rongchang, et al.
Pubblicazione: (2024)

Jailbreaking Commercial Black-Box LLMs with Explicitly Harmful Prompts
di: Zhang, Chiyu, et al.
Pubblicazione: (2025)

Fingerprint Vector: Enabling Scalable and Efficient Model Fingerprint Transfer via Vector Addition
di: Xu, Zhenhua, et al.
Pubblicazione: (2024)

Understanding and Enhancing the Transferability of Jailbreaking Attacks
di: Lin, Runqi, et al.
Pubblicazione: (2025)

HarmRLVR: Weaponizing Verifiable Rewards for Harmful LLM Alignment
di: Liu, Yuexiao, et al.
Pubblicazione: (2025)

The Trojan Knowledge: Bypassing Commercial LLM Guardrails via Harmless Prompt Weaving and Adaptive Tree Search
di: Wei, Rongzhe, et al.
Pubblicazione: (2025)

KinGuard: Hierarchical Kinship-Aware Fingerprinting to Defend Against Large Language Model Stealing
di: Xu, Zhenhua, et al.
Pubblicazione: (2026)

The Jailbreak Tax: How Useful are Your Jailbreak Outputs?
di: Nikolić, Kristina, et al.
Pubblicazione: (2025)

Harmless Backdoor-based Client-side Watermarking in Federated Learning
di: Luo, Kaijing, et al.
Pubblicazione: (2024)

From Compression to Accountability: Harmless Copyright Protection for Dataset Distillation
di: Liang, Yan, et al.
Pubblicazione: (2026)

Temporal UI State Inconsistency in Desktop GUI Agents: Formalizing and Defending Against TOCTOU Attacks on Computer-Use Agents
di: Xu, Wenpeng
Pubblicazione: (2026)

WalletProbe: A Testing Framework for Browser-based Cryptocurrency Wallet Extensions
di: Hu, Xiaohui, et al.
Pubblicazione: (2025)

BaThe: Defense against the Jailbreak Attack in Multimodal Large Language Models by Treating Harmful Instruction as Backdoor Trigger
di: Chen, Yulin, et al.
Pubblicazione: (2024)

When Safe Models Merge into Danger: Exploiting Latent Vulnerabilities in LLM Fusion
di: Li, Jiaqing, et al.
Pubblicazione: (2026)

Involuntary Jailbreak: On Self-Prompting Attacks
di: Guo, Yangyang, et al.
Pubblicazione: (2025)

LLMGuard: Guarding Against Unsafe LLM Behavior
di: Goyal, Shubh, et al.
Pubblicazione: (2024)

Safe2Harm: Semantic Isomorphism Attacks for Jailbreaking Large Language Models
di: Yang, Fan
Pubblicazione: (2025)

Different Paths to Harmful Compliance: Behavioral Side Effects and Mechanistic Divergence Across LLM Jailbreaks
di: Kabir, Md Rysul, et al.
Pubblicazione: (2026)

Propagating Unsafe Actions in LLM Controlled Multi-Robot Collaboration via Single Robot Compromise
di: Huang, Zhen, et al.
Pubblicazione: (2026)

Beyond Jailbreak: Unveiling Risks in LLM Applications Arising from Blurred Capability Boundaries
di: Zhang, Yunyi, et al.
Pubblicazione: (2025)

Revisiting JBShield: Breaking and Rebuilding Representation-Level Jailbreak Defenses
di: Derya, Kemal, et al.
Pubblicazione: (2026)

Targeted Fuzzing for Unsafe Rust Code: Leveraging Selective Instrumentation
di: Paaßen, David, et al.
Pubblicazione: (2025)

Exploiting Prefix-Tree in Structured Output Interfaces for Enhancing Jailbreak Attacking
di: Li, Yanzeng, et al.
Pubblicazione: (2025)

Ellipsoid Control: A White-list Jailbreak Defense via Benign Latent Modeling
di: Chen, Luoyu, et al.
Pubblicazione: (2026)