Copertina

Efficient LLM Moderation with Multi-Layer Latent Prototypes

Fuente: arXiv

Salvato in:

Dettagli Bibliografici
Autori principali:	Chrabąszcz, Maciej, Szatkowski, Filip, Wójcik, Bartosz, Dubiński, Jan, Trzciński, Tomasz, Cygert, Sebastian
Natura:	Preprint
Pubblicazione:	2025
Soggetti:	Machine Learning Artificial Intelligence Computation and Language Cryptography and Security
Accesso online:	Acceder al recurso
Tags:	Aggiungi Tag Nessun Tag, puoi essere il primo ad aggiungerne!!

Documenti analoghi

Monitoring the Internal Monologue: Probe Trajectories Reveal Reasoning Dynamics
di: Chrabąszcz, Maciej, et al.
Pubblicazione: (2026)

Backdoor Vectors: a Task Arithmetic View on Backdoor Attacks and Defenses
di: Pawlak, Stanisław, et al.
Pubblicazione: (2025)

Conditional misalignment: common interventions can hide emergent misalignment behind contextual triggers
di: Dubiński, Jan, et al.
Pubblicazione: (2026)

REALISTA: Realistic Latent Adversarial Attacks that Elicit LLM Hallucinations
di: Liang, Buyun, et al.
Pubblicazione: (2026)

FLAME: Flexible LLM-Assisted Moderation Engine
di: Bakulin, Ivan, et al.
Pubblicazione: (2025)

Addressing The Devastating Effects Of Single-Task Data Poisoning In Exemplar-Free Continual Learning
di: Pawlak, Stanisław, et al.
Pubblicazione: (2025)

Virus: Harmful Fine-tuning Attack for Large Language Models Bypassing Guardrail Moderation
di: Huang, Tiansheng, et al.
Pubblicazione: (2025)

Graph of Attacks with Pruning: Optimizing Stealthy Jailbreak Prompt Generation for Enhanced LLM Content Moderation
di: Schwartz, Daniel, et al.
Pubblicazione: (2025)

Probing the Robustness of Large Language Models Safety to Latent Perturbations
di: Gu, Tianle, et al.
Pubblicazione: (2025)

Tracing the Dynamics of Refusal: Exploiting Latent Refusal Trajectories for Robust Jailbreak Detection
di: Hu, Xulin, et al.
Pubblicazione: (2026)

Latent Fusion Jailbreak: Blending Harmful and Harmless Representations to Elicit Unsafe LLM Outputs
di: Xing, Wenpeng, et al.
Pubblicazione: (2025)

PromptScreen: Efficient Jailbreak Mitigation Using Semantic Linear Classification in a Multi-Staged Pipeline
di: Rao, Akshaj Prashanth, et al.
Pubblicazione: (2025)

Adversarial Intent is a Latent Variable: Stateful Trust Inference for Securing Multimodal Agentic RAG
di: Singh, Inderjeet, et al.
Pubblicazione: (2026)

LLM in the Shell: Generative Honeypots
di: Sladić, Muris, et al.
Pubblicazione: (2023)

Federated In-Context LLM Agent Learning
di: Wu, Panlong, et al.
Pubblicazione: (2024)

AdvPrefix: An Objective for Nuanced LLM Jailbreaks
di: Zhu, Sicheng, et al.
Pubblicazione: (2024)

Policy-Invisible Violations in LLM-Based Agents
di: Wu, Jie, et al.
Pubblicazione: (2026)

Certifying LLM Safety against Adversarial Prompting
di: Kumar, Aounon, et al.
Pubblicazione: (2023)

Beyond Gradient and Priors in Privacy Attacks: Leveraging Pooler Layer Inputs of Language Models in Federated Learning
di: Li, Jianwei, et al.
Pubblicazione: (2023)

Covert Malicious Finetuning: Challenges in Safeguarding LLM Adaptation
di: Halawi, Danny, et al.
Pubblicazione: (2024)

Adaptive Instruction Composition for Automated LLM Red-Teaming
di: Zymet, Jesse, et al.
Pubblicazione: (2026)

LLM Ghostbusters: Surgical Hallucination Suppression via Adaptive Unlearning
di: Spracklen, Joseph, et al.
Pubblicazione: (2026)

SECA: Semantically Equivalent and Coherent Attacks for Eliciting LLM Hallucinations
di: Liang, Buyun, et al.
Pubblicazione: (2025)

Systematically Analyzing Prompt Injection Vulnerabilities in Diverse LLM Architectures
di: Benjamin, Victoria, et al.
Pubblicazione: (2024)

Instructional Segment Embedding: Improving LLM Safety with Instruction Hierarchy
di: Wu, Tong, et al.
Pubblicazione: (2024)

LLM Cyber Evaluations Don't Capture Real-World Risk
di: Lukošiūtė, Kamilė, et al.
Pubblicazione: (2025)

BadAgent: Inserting and Activating Backdoor Attacks in LLM Agents
di: Wang, Yifei, et al.
Pubblicazione: (2024)

Unlearned but Not Forgotten: Data Extraction after Exact Unlearning in LLM
di: Wu, Xiaoyu, et al.
Pubblicazione: (2025)

Capability-Based Scaling Trends for LLM-Based Red-Teaming
di: Panfilov, Alexander, et al.
Pubblicazione: (2025)

AgentSOC: A Multi-Layer Agentic AI Framework for Security Operations Automation
di: Roy, Joyjit, et al.
Pubblicazione: (2026)

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

SELF: A Robust Singular Value and Eigenvalue Approach for LLM Fingerprinting
di: Zhang, Hanxiu, et al.
Pubblicazione: (2025)

A Generative Approach to LLM Harmfulness Mitigation with Red Flag Tokens
di: Dobre, David, et al.
Pubblicazione: (2025)

From Domains to Instances: Dual-Granularity Data Synthesis for LLM Unlearning
di: Xu, Xiaoyu, et al.
Pubblicazione: (2026)

Defending Against Alignment-Breaking Attacks via Robustly Aligned LLM
di: Cao, Bochuan, et al.
Pubblicazione: (2023)

A Fast, Reliable, and Secure Programming Language for LLM Agents with Code Actions
di: Mell, Stephen, et al.
Pubblicazione: (2025)

Tuning without Peeking: Provable Generalization Bounds and Robust LLM Post-Training
di: Labiad, Ismail, et al.
Pubblicazione: (2025)

Measuring Real-World Prompt Injection Attacks in LLM-based Resume Screening
di: Zhang, Mohan, et al.
Pubblicazione: (2026)

Cheating Automatic LLM Benchmarks: Null Models Achieve High Win Rates
di: Zheng, Xiaosen, et al.
Pubblicazione: (2024)

There Are No Silly Questions: Evaluation of Offline LLM Capabilities from a Turkish Perspective
di: Yilmaz, Edibe, et al.
Pubblicazione: (2026)