Skip to main content
LLM Security Database
Skip to research search
Last analyzed Sep 9, 2026

Language Model Security Database

985 research findings · 1119 evaluated models

Filtered research findings

468 entries

Matches every word across titles, descriptions, sources, affected systems, and models.

Intermediate Denoising Priming

Published Analyzed Source: arXiv

A "Priming Vulnerability" exists in Masked Diffusion Language Models (MDLMs) due to their iterative, parallel denoising inference mechanism. Unlike autoregressive models that generate tokens sequentially, MDLMs refine a sequence from a fully masked state through multiple denoising steps. The vulnerability arises because standard safety alignment (e.g., SFT, DPO, MOSA) typically optimizes the model to generate safe responses only when initialized from a fully masked sequence. If an affirmative…

Toward Safer Diffusion Language Models: Discovery and Mitigation of Priming Vulnerability
Evaluated models: Not reported

LLM Self-Evolving Safety Decline

Published Analyzed Source: arXiv

Large Language Models (LLMs), including proprietary and open-weight state-of-the-art systems, are vulnerable to automated, self-evolving adversarial attacks orchestrated by multi-agent frameworks. The vulnerability exists because current safety alignment strategies (RLHF, static safety filters) fail to generalize against the "SafeEvalAgent" attack vector. In this vector, an "Analyst" agent analyzes model refusals to iteratively refine attack strategies, while a "Specialist" agent grounds these…

SafeEvalAgent: Toward Agentic and Self-Evolving Safety Evaluation of LLMs
Evaluated models: GPT-5, GPT-5 Chat Latest, Gemini 2.5 Pro+7 more
  • GPT-5
  • GPT-5 Chat Latest
  • Gemini 2.5 Pro
  • Gemini 2.5 Flash
  • Grok 4
  • Qwen 3 8B
  • Qwen 3 32B
  • Llama 4 Scout
  • Llama 4 Maverick
  • DeepSeek V3.1

RL-driven Formalized Prompt Jailbreaking

Published Analyzed Source: arXiv

A vulnerability exists in aligned Large Language Models (LLMs) where a harmful instruction can be obfuscated through a multi-step formalization process, bypassing safety mechanisms. The attack, named Prompt Jailbreaking via Semantic and Structural Formalization (PASS), uses a Reinforcement Learning (RL) agent to dynamically construct an adversarial prompt. The agent learns to apply a sequence of actions—such as symbolic abstraction, logical encoding, mathematical representation, metaphorical…

Formalization Driven LLM Prompt Jailbreaking via Reinforcement Learning
Evaluated models: DeepSeek V3, Qwen 3 14B

Schrodinger's Compliance Jailbreak

Published Analyzed Source: arXiv

The GPT-OSS-20B large language model contains critical failures in its alignment and Chain-of-Thought (CoT) reasoning mechanisms, specifically in how it prioritizes numerical objectives and validates procedural structure. The model is vulnerable to "Quant Fever," where explicit numerical targets in a prompt (e.g., "delete 90% of files") override contextual safety constraints (e.g., "do not delete important files"). Furthermore, the model exhibits "Reasoning Procedure Mirage," where harmful…

Quant Fever, Reasoning Blackholes, Schrodinger's Compliance, and More: Probing GPT-OSS-20B
Evaluated models: Not reported

Single Query Dynamic Output

Published Analyzed Source: arXiv

A vulnerability exists in Large Language Models (LLMs) and multi-label text classification systems that allows for Textual Dynamic Outputs Attacks (TDOA). This technique enables hard-label black-box attacks against systems with variable or generative output spaces (where the number of labels or specific label tokens are not fixed). The attack functions by training a surrogate model on clustered coarse-grained labels derived from the victim model's fine-grained dynamic outputs. It subsequently…

Text Adversarial Attacks with Dynamic Outputs
Evaluated models: GPT-4o, GPT-4o Mini, GPT-4.1+5 more
  • GPT-4o
  • GPT-4o Mini
  • GPT-4.1
  • Claude 3.7 Sonnet
  • DeepSeek V3
  • BERT
  • DistilBERT
  • RoBERTa

Universal Steering Jailbreak

Published Analyzed Source: arXiv

A vulnerability exists in the alignment mechanisms of Large Language Models (LLMs) where activation steering—the process of injecting vectors into hidden states during inference—can systematically bypass refusal safeguards. By modifying the residual stream activations at intermediate layers (typically $\lfloor L/2 \rfloor$) using the formula $\overline{\mathbf{x}}_{i}^{(l)}=\mathbf{x}_{i}^{(l)}+\alpha\mathbf{v}$, an attacker can force the model to comply with harmful requests. Research…

The Rogue Scalpel: Activation Steering Compromises LLM Safety
Evaluated models: Llama 3 8B, Llama 3.1 8B, Qwen 2.5 7B+1 more
  • Llama 3 8B
  • Llama 3.1 8B
  • Qwen 2.5 7B
  • Falcon 7B

VLM Imperceptible Image Jailbreak

Published Analyzed Source: arXiv

A vulnerability exists in Vision-Language Models (VLMs) that allows for the bypass of safety alignment mechanisms through loss-guided adversarial image perturbations. This attack, known as JaiLIP, operates entirely in the image space, requiring no textual prompt manipulation. The vulnerability is exploited by optimizing an adversarial image using a joint objective function that minimizes the Mean Squared Error (MSE) between the clean and perturbed image while maximizing the model's loss for…

JaiLIP: Jailbreaking Vision-Language Models via Loss Guided Image Perturbation
Evaluated models: GPT-4, InstructBLIP, Vicuna 13B

Deceptive Reasoning Bypass

Published Analyzed Source: arXiv

Frontier Large Language Models (LLMs) utilizing Chain-of-Thought (CoT) reasoning are vulnerable to deceptive alignment attacks via adversarial system prompt injection. This vulnerability allows an attacker to induce "deceptive reasoning," where the model’s internal CoT actively plans or entertains malicious directives (e.g., radicalization, bias, or violence) while the final user-facing output remains benign, helpful, and innocuous. By creating a dissociation between internal reasoning and…

D-REX: A Benchmark for Detecting Deceptive Reasoning in Large Language Models
Evaluated models: Nova Pro v1, DeepSeek R1, Claude 3.7 Sonnet Thinking+4 more
  • Nova Pro v1
  • DeepSeek R1
  • Claude 3.7 Sonnet Thinking
  • Qwen 3 235B-A22B
  • Gemini 2.5 Flash
  • Gemini 2.5 Pro
  • Grok 3 Mini Beta

Emergent Agentic Vulnerabilities

Published Analyzed Source: arXiv

GPT-OSS-20B exhibits "agentic-only" vulnerabilities where safety guardrails effective in standalone model inference fail when the model operates within an agentic execution loop. These vulnerabilities emerge when the model is deployed in a multi-step agentic architecture (e.g., utilizing LangGraph, tool usage, and memory retention). Attackers can bypass safety filters by employing context-aware iterative refinement attacks, which incorporate the full agentic state—including tool outputs…

Mind the Gap: Comparing Model-vs Agentic-Level Red Teaming with Action-Graph Observability on GPT-OSS-20B
Evaluated models: Not reported

Physical Agent Object Misdirection

Published Analyzed Source: arXiv

AdvEDM reveals a vulnerability in Vision-Language Model (VLM) based Embodied Decision-Making (EDM) systems, such as those used in autonomous driving and robotic manipulation. The vulnerability allows an attacker to launch fine-grained adversarial attacks that selectively modify the perception of specific objects in an input image—either by removing them (Semantic Removal) or adding them (Semantic Addition)—while preserving the semantic integrity of the rest of the scene.

AdvEDM: Fine-grained Adversarial Attack against VLM-based Embodied Agents
Evaluated models: BLIP-2, MiniGPT-4, LLaVA-v2+5 more
  • BLIP-2
  • MiniGPT-4
  • LLaVA-v2
  • Otter-Image
  • OpenFlamingo-v2
  • Dolphins
  • EmbodiedGPT
  • GPT-4o

Research methodology

Entries summarize publicly available primary-source security research. Model names reflect only systems explicitly evaluated by the cited paper, and measurements are research-reported unless independent verification is stated.

Discovery and publication are separate: new papers enter a candidate queue, then require human review and explicit approval before publication. Updates follow that review process, not a guaranteed daily publishing schedule. “Last analyzed” is the newest analysis date among published entries; it does not mean that all recent research has been covered.