Cover Image

Do Prompts Guarantee Safety? Mitigating Toxicity from LLM Generations through Subspace Intervention

Fuente: arXiv

Saved in:

Bibliographic Details
Main Authors:	Singh, Himanshu, Xu, Ziwei, Subramanyam, A. V., Kankanhalli, Mohan
Format:	Preprint
Published:	2026
Subjects:	Computation and Language Cryptography and Security
Online Access:	Acceder al recurso
Tags:	Add Tag No Tags, Be the first to tag this record!

Similar Items

Relevance as a Vulnerability: How Web Retrieval Degrades Safety Alignment in LLM Agents
by: Nawal, Aditya, et al.
Published: (2026)

LLMs Can Unlearn Refusal with Only 1,000 Benign Samples
by: Guo, Yangyang, et al.
Published: (2026)

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

The VLLM Safety Paradox: Dual Ease in Jailbreak Attack and Defense
by: Guo, Yangyang, et al.
Published: (2024)

Preference Tuning For Toxicity Mitigation Generalizes Across Languages
by: Li, Xiaochen, et al.
Published: (2024)

Mitigating Fine-tuning based Jailbreak Attack with Backdoor Enhanced Safety Alignment
by: Wang, Jiongxiao, et al.
Published: (2024)

Understanding and Mitigating Over-refusal for Large Language Models via Safety Representation
by: Zhang, Junbo, et al.
Published: (2025)

Safety is Not Only About Refusal: Reasoning-Enhanced Fine-tuning for Interpretable LLM Safety
by: Zhang, Yuyou, et al.
Published: (2025)

T2ISafety: Benchmark for Assessing Fairness, Toxicity, and Privacy in Image Generation
by: Li, Lijun, et al.
Published: (2025)

Prompt Optimization and Evaluation for LLM Automated Red Teaming
by: Freenor, Michael, et al.
Published: (2025)

Confidential Prompting: Privacy-preserving LLM Inference on Cloud
by: Li, Caihua, et al.
Published: (2024)

Raccoon: Prompt Extraction Benchmark of LLM-Integrated Applications
by: Wang, Junlin, et al.
Published: (2024)

DPTraj-PM: Differentially Private Trajectory Synthesis Using Prefix Tree and Markov Process
by: Wang, Nana, et al.
Published: (2024)

GradSafe: Detecting Jailbreak Prompts for LLMs via Safety-Critical Gradient Analysis
by: Xie, Yueqi, et al.
Published: (2024)

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

Jailbreaking LLMs via Semantically Relevant Nested Scenarios with Targeted Toxic Knowledge
by: Xu, Ning, et al.
Published: (2025)

Mitigating the Safety-utility Trade-off in LLM Alignment via Adaptive Safe Context Learning
by: Wang, Yanbo, et al.
Published: (2026)

Root Defence Strategies: Ensuring Safety of LLM at the Decoding Level
by: Zeng, Xinyi, et al.
Published: (2024)

Self-Disguise Attack: Induce the LLM to disguise itself for AIGT detection evasion
by: Zhou, Yinghan, et al.
Published: (2025)

The Landscape of Prompt Injection Threats in LLM Agents: From Taxonomy to Analysis
by: Wang, Peiran, et al.
Published: (2026)

GuidedBench: Measuring and Mitigating the Evaluation Discrepancies of In-the-wild LLM Jailbreak Methods
by: Huang, Ruixuan, et al.
Published: (2025)

Good Parenting is all you need -- Multi-agentic LLM Hallucination Mitigation
by: Kwartler, Ted, et al.
Published: (2024)

PandaGuard: Systematic Evaluation of LLM Safety against Jailbreaking Attacks
by: Shen, Guobin, et al.
Published: (2025)

Adversarial Attacks on LLM-as-a-Judge Systems: Insights from Prompt Injections
by: Maloyan, Narek, et al.
Published: (2025)

PSM: Prompt Sensitivity Minimization via LLM-Guided Black-Box Optimization
by: Jawad, Huseein, et al.
Published: (2025)

TWGuard: A Case Study of LLM Safety Guardrails for Localized Linguistic Contexts
by: Chu, Hua-Rong, et al.
Published: (2026)

One Word at a Time: Incremental Completion Decomposition Breaks LLM Safety
by: Arif, Samee, et al.
Published: (2026)

SecureForge: Finding and Preventing Vulnerabilities in LLM-Generated Code via Prompt Optimization
by: Liu, Houjun, et al.
Published: (2026)

Technical Report for ICML 2024 TiFA Workshop MLLM Attack Challenge: Suffix Injection and Projected Gradient Descent Can Easily Fool An MLLM
by: Guo, Yangyang, et al.
Published: (2024)

SafeReview: Defending LLM-based Review Systems Against Adversarial Hidden Prompts
by: Xin, Yuan, et al.
Published: (2026)

Model-Agnostic Lifelong LLM Safety via Externalized Attack-Defense Co-Evolution
by: Zhang, Xiaozhe, et al.
Published: (2026)

SecurityLingua: Efficient Defense of LLM Jailbreak Attacks via Security-Aware Prompt Compression
by: Li, Yucheng, et al.
Published: (2025)

SoK: Privacy Risks and Mitigations in Retrieval-Augmented Generation Systems
by: Bodea, Andreea-Elena, et al.
Published: (2026)

How Alignment and Jailbreak Work: Explain LLM Safety through Intermediate Hidden States
by: Zhou, Zhenhong, et al.
Published: (2024)

RAG Safety: Exploring Knowledge Poisoning Attacks to Retrieval-Augmented Generation
by: Zhao, Tianzhe, et al.
Published: (2025)

FFT: Towards Harmlessness Evaluation and Analysis for LLMs with Factuality, Fairness, Toxicity
by: Cui, Shiyao, et al.
Published: (2023)

One Model Transfer to All: On Robust Jailbreak Prompts Generation against LLMs
by: Li, Linbao, et al.
Published: (2025)

BiAxisAudit: A Novel Framework to Evaluate LLM Bias Across Prompt Sensitivity and Response-Layer Divergence
by: Gan, Jialing, et al.
Published: (2026)

Subspace Defense: Discarding Adversarial Perturbations by Learning a Subspace for Clean Signals
by: Zheng, Rui, et al.
Published: (2024)

Efficient Detection of Toxic Prompts in Large Language Models
by: Liu, Yi, et al.
Published: (2024)