Attack Context

White-box Research

Attacks leveraging model implementation details

11 matching entries out of 228 in this category

White-box research results

11 entries

LLM Judge Manipulation

Published Mar 31, 2026Analyzed Apr 10, 2026Source: arXiv

Generative reward models deployed as LLM-as-a-Judge (LaaJ) evaluators contain a logic bypass vulnerability where superficial "master key" inputs trigger false positive rewards regardless of actual response quality. Instead of evaluating the candidate's output, large judge models are inadvertently triggered by specific token sequences to solve the prompt independently. This allows malicious actors or policy models undergoing reinforcement learning to consistently game the reward signal by…

Security in LLM-as-a-Judge: A Comprehensive SoK

Evaluated models: GPT-4o, o1, Qwen 2.5 72B Instruct+1 more

GPT-4o
o1
Qwen 2.5 72B Instruct
Llama 3 70B Instruct

Covert Grade Manipulation

Published Feb 1, 2026Analyzed Feb 22, 2026Source: arXiv

Large Language Models (LLMs) utilized for Automatic Short Answer Grading (ASAG) are vulnerable to the "GradingAttack" framework, which employs fine-grained adversarial manipulation to alter grading outcomes. Attackers can leverage two distinct strategies: (1) Prompt-level attacks using role-play injection strings that instruct the model to pretend an answer is correct regardless of factual accuracy, and (2) Token-level attacks utilizing gradient-based optimization (similar to Greedy Coordinate…

GradingAttack: Attacking Large Language Models Towards Short Answer Grading Ability

Evaluated models: GPT-3.5, GPT-4, GPT-4o+3 more

GPT-3.5
GPT-4
GPT-4o
Llama 3.1 8B
Mistral 7B
Qwen 2.5 7B

Unified Robustness Gap

Published Jan 31, 2026Analyzed Mar 9, 2026Source: arXiv

Text scoring models, including dense retrievers, rerankers, and reward models, are vulnerable to score manipulation attacks via search-based discrete perturbations and content injection. An attacker can systematically modify candidate texts using rudimentary string manipulations, gradient-guided token swaps (e.g., HotFlip), masked language modeling (MLM) swaps, or query/sentence injections to spuriously increase model scores. This structural failure condition allows an irrelevant passage or a…

Unifying Adversarial Robustness and Training Across Text Scoring Models

Evaluated models: E5 BERT-base, Qwen 3 0.6B, Llama 3.2 3B Instruct+2 more

E5 BERT-base
Qwen 3 0.6B
Llama 3.2 3B Instruct
Llama 3.1 8B Instruct
Skywork Reward V2

LLM Router Rerouting

Published Jan 29, 2026Analyzed Mar 8, 2026Source: arXiv

LLM routing systems are vulnerable to adversarial rerouting attacks where malicious triggers prepended to user queries manipulate the router's model-selection mechanism. Because LLM routers function as classifiers evaluating query complexity to balance computational cost and response quality, an attacker can craft adversarial prefixes that distort the query's latent semantic representation. This exploits the router's decision boundaries, forcing the system to misclassify the input and redirect…

RerouteGuard: Understanding and Mitigating Adversarial Risks for LLM Routing

Evaluated models: GPT-4, GPT-4o, GPT-5+2 more

GPT-4
GPT-4o
GPT-5
Llama 3 8B
Mixtral 8x7B

Embodied Cross-Modal Misalignment

Published Nov 20, 2025Analyzed Jan 14, 2026Source: arXiv

OpenVLA, a Vision-Language-Action (VLA) model, contains a vulnerability regarding multimodal adversarial robustness. The model lacks sufficient cross-modal alignment stability, allowing attackers to disrupt the grounding between visual perception and linguistic instructions. By utilizing the "VLA-Fool" framework, adversaries can inject perturbations via three vectors: (1) Semantically Greedy Coordinate Gradient (SGCG), which alters specific linguistic tokens (referential cues, attributes…

When alignment fails: Multimodal adversarial attacks on vision-language-action models

Evaluated models: Not reported

Schrodinger's Compliance Jailbreak

Published Sep 28, 2025Analyzed Feb 21, 2026Source: 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

Copy-Guided Reasoning Corruption

Published Jul 22, 2025Analyzed Dec 30, 2025Source: arXiv

Reasoning-capable Large Language Models (LLMs) are vulnerable to a class of indirect prompt injection known as Copy-Guided Attacks (CGA). This vulnerability exploits the intrinsic behavior of reasoning models to copy tokens from the input prompt (such as variable names, function identifiers, or code snippets) into their intermediate reasoning traces (Chain-of-Thought). By embedding adversarial trigger sequences into external payloads—specifically within data the model is expected to analyze—an…

When LLMs Copy to Think: Uncovering Copy-Guided Attacks in Reasoning LLMs

Evaluated models: DeepSeek R1 Distill Qwen 1.5B, DeepSeek R1 Distill Llama 8B

Airborne Audio LLM Hijack

Published Jul 7, 2025Analyzed Jan 14, 2026Source: arXiv

Audio-based Large Language Models (ALLMs), specifically Qwen2-Audio, are vulnerable to over-the-air adversarial audio attacks. An attacker with white-box access can generate robust adversarial audio perturbations using gradient-based optimization combined with audio augmentation techniques (specifically SpecAugment, translation, and additive noise). These perturbations, when played through a speaker in the physical environment, manipulate the ALLM processing the audio via a microphone. This…

Attacker's Noise Can Manipulate Your Audio-based LLM in the Real World

Evaluated models: Not reported

LLM Judge Subversion

Published Jun 11, 2025Analyzed Dec 9, 2025Source: arXiv

Alibaba Cloud PAI-Judge and PAI-Judge-Plus are vulnerable to a composite adversarial attack that exploits attention mechanism limitations in Large Language Models (LLMs). An authenticated attacker can manipulate automated evaluation outcomes by appending a long, irrelevant text suffix (approximately 1000 to 2000+ characters) to a response containing adversarial perturbations. This "long-suffix" strategy overwhelms the judge model's context window, causing the attention mechanism to degrade and…

LLMs Cannot Reliably Judge (Yet?): A Comprehensive Assessment on the Robustness of LLM-as-a-Judge

Evaluated models: GPT-4o, Llama 3.1 8B, Llama 3.3 70B+3 more

GPT-4o
Llama 3.1 8B
Llama 3.3 70B
Mistral 7B
DeepSeek R1
Qwen 2.5 7B

JailFlip Implicit Harm

Published Jun 9, 2025Analyzed Dec 8, 2025Source: arXiv

A vulnerability exists in the safety alignment mechanisms of Large Language Models (LLMs) (including GPT-4, Claude 3, Gemini, and Qwen families) leading to "Implicit Harm." Unlike traditional jailbreaks that use overtly harmful queries, this vulnerability allows remote attackers to coerce the model into providing factually incorrect, plausible, and dangerous responses to benign-looking inputs. By employing "JailFlip" techniques—specifically constructed affirmative-type or denial-type queries…

Beyond Jailbreaks: Revealing Stealthier and Broader LLM Security Risks Stemming from Alignment Failures

Evaluated models: GPT-4.1, GPT-4.1 Mini, GPT-4o+3 more

GPT-4.1
GPT-4.1 Mini
GPT-4o
GPT-4o Mini
Qwen Plus
Qwen Turbo

Related research categories